Weekend Notebook #2621 – Google I/O 2026 and Europe AI Acceleration

amitabh apte

Published on LinkedIn and amitabhapte.com on 24th May 2026

Three stories this week. Google reminded the market it has structural advantages no challenger can easily replicate. European infrastructure stocks confirmed AI capex is now a global wealth story, not a West Coast one. And a Formula 1 team alongside a decade of productivity research told the honest story about what AI can and cannot yet do in production.

1. Google Fights Back, and the Numbers Are Serious

Google I/O 2026 was not a product showcase. It was a statement of scale. Google’s models now process 3.2 quadrillion tokens per month, up 7x from last year. AI Overviews has 2.5 billion monthly users. AI Mode in Search crossed 1 billion monthly active users in just one year. Thirteen Google products each have over a billion users. No AI challenger has a distribution surface remotely close to this.

The headline product was Gemini Spark, a 24/7 personal AI agent that runs in the background on Google Cloud, continuing tasks even when your phone is locked. It integrates with Gmail, Docs, Calendar, and third-party apps. Users teach it recurring workflows: scan bills monthly, generate documents from meeting notes, draft follow-up emails. The model underneath, Gemini 3.5 Flash, now outperforms the previous flagship on agentic benchmarks at four times the speed of GPT-5.5 and Claude Opus 4.7.

The Economist put it plainly: Google is dethroning OpenAI as the king of consumer AI. More people now download Gemini than ChatGPT. Search, Android, Chrome, Gmail: the distribution advantage is not just a moat, it is a compounding asset. Pichai was also notably candid in a post-I/O interview, saying he understands why people are anxious about AI and is not dismissing it. Measured rather than triumphant, in a week of substantial announcements.

My PoV: OpenAI built the category. Google is using 25 years of distribution infrastructure to absorb it. The practical implication for enterprise leaders: Gemini Spark’s deep Workspace integration means the AI agent most likely to reach your workforce at scale may not arrive through a procurement decision. It will arrive through the productivity suite you already pay for. That changes the governance conversation significantly.

2. The AI Wealth Is Spreading to Europe

A quieter story has been building in Europe. This week it became hard to ignore. Aixtron is up 189% year-to-date, Technoprobe 129%, STMicroelectronics 133%, Nokia 108%. Nokia, written off by most investors as a legacy phone maker, has repositioned as an AI networking infrastructure provider. Its AI and cloud infrastructure revenue grew 49% in Q1 2026. JP Morgan’s framing is precise: “In Europe, scarcity amplifies the trend. There are few large, liquid AI pure-plays, so flows concentrate in a small group of perceived AI proxies.”

But the returns are not sentiment alone. These companies make equipment that goes into every AI data centre being built globally: compound semiconductor deposition tools, optical fibre networking, chip testing rigs, power semiconductors for data centre power management. The Stoxx Europe Semiconductor index is up 74% in 2026, against 2% for the broad Stoxx Europe 600. That divergence inside a flat market tells the structural story. The picks and shovels are as European as they are American.

My PoV: Nokia’s transformation from legacy telecom to AI networking player is a case study in what a decisive infrastructure pivot can do to a company’s market position. The question for technology leaders is not which European stocks to own. It is which of your infrastructure and technology partners are making a similar transition, and whether your roadmap accounts for their success or failure in doing so.

3. AI in Production: The Honest Numbers

Ferrari and IBM opened their playbook to TechCrunch this week. The Ferrari app, rebuilt on IBM watsonx, converts millions of race telemetry data points into personalised narratives for 400 million Tifosi worldwide. Results: 62% increase in engagement over race weekends, 56% more race-active users, 35% more time in app. The goal is not broadcasting at fans. It is making each of them feel known. AI makes personalisation at 400 million people simultaneously possible. No human content team could do that.

The FT’s productivity analysis offered the necessary counterweight. A study by nonprofit METR found AI tools made software developers’ tasks take 20% longer. An NBER survey of thousands of executives found negligible measured productivity impact in 2025. UC Berkeley researchers found AI-assisted workers took on more tasks but did more overall work, with multitasking driving cognitive fatigue rather than efficiency. The gap between the productivity AI promises and what organisations are measuring remains wide.

My PoV: Both stories are true simultaneously. Ferrari’s results are real where the use case is specific, the data is clean, and the workflow is redesigned around AI. The productivity paradox persists where AI is added to existing processes without changing them. The tool gets deployed. The process does not change. That distinction is the most useful frame I have seen this year for separating AI deployments that will deliver from those that will not.

My Takeaway This Weekend

Distribution at scale is a different kind of moat than model capability. Gemini Spark running behind a billion Gmail inboxes is a more durable competitive position than any benchmark score. Europe’s infrastructure rally confirms the AI buildout is global. And the Ferrari versus productivity research contrast is the clearest lens available for evaluating your own AI deployments. The difference is not the technology. It is whether the process was actually redesigned to use it.

Weekend Notebook #2620 – when the Stack goes Public

amitabh apte

Published on LinkedIn and amitabhapte.com  |  17 May 2026

This week, three very different stories arrived at the same conclusion. A chipmaker went public and nearly doubled on its first day. India signed a deal to build its first semiconductor fab. And AI-powered robots quietly took over 40 Los Angeles neighbourhoods while AI rewired the world’s largest e-commerce search bar. The infrastructure of intelligence is no longer being planned. It is being built, listed, traded, and deployed on your nearest pavement.

1. The Capital Signal: Chips, Memory and Public Markets

The Cerebras Systems IPO was the largest US tech listing since Uber in 2019. Priced at $185 per share, it raised $5.55 billion and surged 68% on its first day of trading, valuing the company at nearly $50 billion. The company makes wafer-scale AI chips. Its WSE-3 processor is the size of a dinner plate, packing over 4 trillion transistors, and is purpose-built for inference, running trained AI models fast and cheaply rather than training them. With AWS and OpenAI already as anchor customers, the order book closed 20 times oversubscribed.

The headline, though, is not just the company. It is the queue behind it. SpaceX, OpenAI, and Anthropic are all reportedly preparing listings, with SpaceX and OpenAI alone expected to raise a combined $135 billion. The public markets are signalling something clear: AI infrastructure is now a provable, priceable asset class, not a speculative bet.

And yet, as capital celebrates the chip layer, a different bottleneck is quietly tightening. High-bandwidth memory (HBM) has become AI’s most critical constraint. Data centres now consume an estimated 70% of global high-end memory production. Samsung, SK Hynix, and Micron have reallocated the vast majority of their fab capacity to HBM for AI accelerators, producing 3x less conventional DRAM per wafer as a result. DRAM prices have surged sharply in early 2026, with memory sold out across the supply chain into 2027. The irony is structural: we are celebrating the intelligence layer while quietly starving the memory layer that makes it run.

My PoV The Cerebras IPO is important, but memory is the signal that matters more. When three manufacturers control 95% of global DRAM production and have committed most of it to AI data centres through 2027, every enterprise architecture conversation should include a memory procurement question. If your AI roadmap does not account for infrastructure scarcity, it is optimistic by design.

2. Geography as Strategy: Capital, Compute and Sovereignty

The most telling signal this week was not a product launch. It was a portfolio filing. A Reuters analysis of nearly 6,000 institutional investor 13-F filings showed that more than 4,000 funds added to or initiated positions in AI infrastructure stocks in Q1 2026. Only 146 sold. Not only that: there were no net sellers of utility stocks in the quarter, as pension funds and endowments moved money toward the power and cooling layer that keeps AI running. The Magnificent Seven are being treated selectively. The infrastructure beneath them is being bought without hesitation.

This capital movement is not random. It is following a geography. AWS committed €33 billion to Spain, framing the country as its European AI epicentre. Tata Electronics and ASML signed an MoU to build India’s first commercial 300mm semiconductor fab in Dholera, Gujarat, backed by an $11 billion investment. The deal, signed in the presence of both nations’ heads of government, covers ASML’s full lithography suite, talent development, and supply chain resilience, targeting chips for automotive, mobile, and AI from 28nm to 110nm. On services, TCS declaring its ambition to become the world’s largest AI-led technology services company, with 130 of its top 139 enterprise clients already on board and over 270,000 employees upskilled in AI in a single year. TCS Chairman N. Chandrasekaran describing AI as “the infrastructure of intelligence”, a deliberate echo of how the internet was framed in the late 1990s. Silicon, services, and strategy, converging at the same moment.

The broader pattern is clear. AI leadership is being contested not just in model benchmarks but in megawatts, fab nodes, data centre land, and sovereign infrastructure decisions. The race has moved from algorithms to atlases.

My PoV When 4,000 institutions buy AI infrastructure and zero sell utilities in the same quarter, that is not a trend. It is a conviction. The smart money has concluded that the physical layer of AI, power, chips, cooling, connectivity, is where durable advantage compounds. For enterprise technology leaders, the implication is concrete: your AI strategy is increasingly shaped by decisions being made in fabs, power grids, and diplomatic meetings, not just in model releases. Understanding your infrastructure dependencies by geography is no longer optional. 

3. AI Meets Culture: From Cannes to the Pavement

Two very different scenes this week, both telling the same story about normalisation.

At Cannes, filmmakers shifted from resistance to pragmatism. French director Xavier Gens noted that AI would have cut his Netflix hit’s visual effects budget in half and saved eight months of post-production time. Demi Moore, serving as a jury member, said simply: “AI is here. To fight it is to fight something that is a battle we will lose.” The distinction that emerged from Cannes was useful and important: directors broadly agreed that using AI to generate scripts or entire films from a prompt should not be allowed, but its use in production and post-production is increasingly accepted. Guillermo del Toro put it well, noting that lumping all AI applications under one label makes productive discussion impossible. That nuance, creativity vs. craft, authorship vs. tooling, is the conversation every industry will eventually need to have. Cannes just had it in public first.

On the commerce side, Amazon replaced Rufus with Alexa for Shopping, embedding an AI assistant directly into the default search bar of the world’s largest e-commerce platform. No click required. AI answers now appear by default for complex queries, completing the shift from search as retrieval to search as conversation. The assistant can also shop third-party retailers and execute purchases autonomously through Buy for Me. Meanwhile, on the streets of Los Angeles, Serve Robotics has deployed over 500 sidewalk delivery bots across 40 neighbourhoods, up from just two in 2023. Cities are scrambling to draft regulations in real time.

The pattern across Cannes, Amazon, and the pavements of Los Angeles is the same. The debate has moved from “will AI arrive?” to “how do we live with it?” That is a meaningful shift. It means AI is no longer being evaluated. It is being negotiated.

My PoV The Cannes conversation matters for leaders well beyond media. Every industry will face the same negotiation: which parts of creative and knowledge work can be augmented, which parts should remain human, and how do you draw that line without either losing competitive ground or losing your people? The organisations that will navigate this best are those that have the conversation explicitly, rather than letting it resolve through default or drift. 

My Takeaway This Weekend

Three things became measurably more real this week. The chip layer went public and priced at scale. The geographic layer hardened, with capital, fabs, and services converging by region rather than by company. And the consumer layer normalised, quietly and irreversibly, from the search bar to the pavement.

The window between “this is coming” and “this has already happened” keeps closing faster than planning cycles allow. The organisations best positioned are not necessarily those with the most AI projects. They are the ones that understand where their infrastructure dependencies sit geographically, which parts of their work they are willing to delegate, and which parts they are determined to protect.

Weekend Notebook #2619 – When the Chip Race Widens and AI Starts Executing

amitabh apte

Published on LinkedIn and amitabhapte.com on 10th May 2026

Three themes this week. The semiconductor wealth that spent three years concentrating in one company is spreading. Anthropic’s compute hunger is turning unlikely infrastructure players into AI winners. And AI has quietly crossed from advising to executing, in cybersecurity, commerce, and code. The pace of change is not slowing. It is compounding.

1. The Chip Race Widens

Wall Street called it a “changing of the guard in AI.” Intel, AMD, and Micron each gained around 25% or more this week, while Nvidia rose a more modest 8%. Intel has more than doubled year-to-date. Micron has surged over 750% in the past year and crossed an $800 billion market cap. The rotation reflects a maturing thesis: AI infrastructure now requires CPUs, memory, and optical networking at scale, not just GPUs. Memory is in shortage, with Micron’s CEO noting customers are receiving only 50 to 65% of their requirements. That squeeze is driving prices up and turning a 47-year-old company in Idaho into one of the hottest trades in the market.

Nvidia is not standing still. It has committed more than $40 billion in equity investments in 2026 alone, anchored by its $30 billion stake in OpenAI and multi-billion dollar deals with Corning and data centre operator IREN. CEO Jensen Huang’s stated logic: “We don’t pick winners. We need to support everyone.” Critics call it circular. Nvidia invests in customers who buy its chips, who then generate the demand that justifies the investment. But there is a harder long-term pressure building. Hyperscalers are building their own AI chips: Google’s TPUs, Amazon’s Trainium, Meta’s MTIA silicon. Amazon’s custom chip business has already crossed a $20 billion revenue run rate. For inference workloads, purpose-built silicon is faster and cheaper than general-purpose GPUs. Nvidia’s dominance in training is secure for now. Its dominance across the full stack is not.

My PoV: The AI chip story is no longer a single-stock thesis. For technology leaders building infrastructure strategy, the question is no longer just “how much Nvidia capacity can we secure?” It is “where does custom silicon from our cloud provider give us a cost and performance advantage for inference at scale?” That is an architecture decision, and it belongs in your AI infrastructure roadmap now.

2. Compute Scarcity Is Creating Unlikely Winners

Anthropic has experienced 80x growth in annualised revenue and usage in Q1 2026. The constraint is not demand. It is compute. The company is buying capacity from every available source, and this week that search reached an unexpected destination. Akamai, a 28-year-old content delivery network founded at MIT, signed a $1.8 billion, seven-year cloud infrastructure deal with Anthropic. The largest contract in Akamai’s history. Its stock rose 27% on the news. Akamai’s 4,000-plus global network locations, built to deliver web content without latency, are being repurposed to run AI inference at the edge. The company’s cloud infrastructure revenue was already up 40% year-on-year before this deal. The Anthropic commitment gives it revenue visibility its legacy CDN business never offered.

Japan joined the story through SoftBank. The Nikkei reached a record high this week, driven in significant part by SoftBank’s AI positioning. SoftBank has committed $100 billion to AI infrastructure in the US through the Stargate joint venture with Oracle, and its domestic portfolio of AI investments has been re-rated sharply upward. Japan is no longer a spectator in the AI infrastructure cycle. Through SoftBank’s capital and government-backed semiconductor incentives, it is an active participant.

My PoV: Akamai’s transformation is the week’s most instructive story for enterprise technology leaders. A company that spent 25 years building global network infrastructure is now one of Anthropic’s most important compute partners. The lesson: AI’s compute hunger is so acute that infrastructure built for entirely different purposes, latency-optimised content delivery, is being repurposed and re-valued. Your own organisation’s infrastructure assets may have strategic value in the AI economy that your current roadmap does not account for.

3. AI Moves from Advising to Executing

OpenAI launched GPT-5.5-Cyber, a variant of its latest model specifically trained to be more permissive for security workflows, available in limited preview to vetted cybersecurity teams. It assists with vulnerability identification, malware analysis, binary reverse engineering, and patch validation. It does not write malware or steal credentials, but it removes the friction that the standard model imposes on legitimate security work. The approach contrasts with Anthropic’s Mythos: OpenAI is betting on broader, verified access through its Trusted Access for Cyber programme, scaled to thousands of individual defenders and hundreds of teams. Anthropic restricted Mythos to around 40 organisations. Two philosophies on the same underlying problem: how do you democratise AI-assisted defence without arming attackers?

In commerce, Alibaba announced the integration of its Qwen AI platform directly into Taobao, giving the agent access to over four billion products across Taobao and Tmall, plus logistics, after-sales, and Alipay checkout. The shopper asks; the agent browses, compares, applies 30-day price tracking, and completes the transaction. This is the largest agentic commerce deployment yet from any platform globally. Western equivalents, including Amazon’s Rufus and Shopify’s AI integrations, remain in the advisory lane: they help you decide, but you still transact. Alibaba’s design puts the agent in the execution seat end-to-end. That is a different operating model, and it will be studied.

In software, Google’s internal struggle to coordinate its AI coding tools is handing competitive ground to Anthropic’s Claude Code and OpenAI’s Codex. Google has world-class models and one of the largest developer communities in the world. But internal alignment on how to deploy those models for coding has lagged, and the market has noticed. Claude Code, which powers agentic multi-file coding tasks, has become the default tool for serious AI-assisted software development at thousands of enterprises. Execution, not capability, is the differentiator.

My PoV: The thread connecting cybersecurity, agentic commerce, and AI coding is the same: AI is no longer in the advisory seat. It is in the execution seat. For technology and business leaders, this changes the governance question. The relevant question is no longer “is this AI recommendation accurate?” It is “who is accountable when this AI action has consequences?” That accountability framework needs to be designed before the deployment, not after the incident.

My Takeaway This Weekend

The AI infrastructure story is maturing and broadening simultaneously. The semiconductor cycle is distributing wealth beyond the initial GPU monopoly. Compute scarcity is elevating infrastructure companies that were not in the AI conversation two years ago. And AI capability is crossing the line from insight to action across security, commerce, and software development.

Three years into the generative AI era, the competitive advantage is no longer who has access to the best model. It is who has the infrastructure to run it, the governance to deploy it responsibly, and the organisational clarity to act faster than the competition. The architecture decisions being made this quarter will set the boundaries of what is possible in 2027. That window is shorter than most planning cycles assume.

Weekend Notebook #2618 – When Capital meets Consequence

amitabh apte

Published on LinkedIn and amitabhapte.com on 3rd May 2026

April closed with the Nasdaq posting its best month since 2020. Underneath the headline: AI is moving from investment thesis to earnings reality. The infrastructure that sustains it is scaling faster than most plans account for. And the same capabilities lifting productivity are lowering the cost of attack. Three stories. One consistent pressure.

1. The Investment Cycle Starts Paying Out

The Nasdaq gained 15.3% in April. Alphabet rose 34%, Amazon 27%, AMD 74%, Micron 61%. These are not projections. They are quarterly results, grounded in cloud and AI workload growth. Goldman Sachs estimates AI investment will drive 40% of S&P 500 earnings-per-share growth this year. The cycle is no longer speculative.

The platform layer is being restructured at the same time. OpenAI ended its exclusive arrangement with Microsoft, capping the revenue share and freeing its models to deploy across any cloud. Within 24 hours, a major AWS partnership was announced. OpenAI’s revenue chief had stated internally that the Microsoft deal had “limited our ability to meet enterprises where they are.” Microsoft retains the primary relationship and IP licence through 2032. But OpenAI is now a multi-cloud business. The cloud competition for AI workloads has reopened.

Meanwhile, Anthropic is in discussions to raise $50 billion at a $900 billion valuation, which would surpass OpenAI’s most recent post-money figure and likely be its final private round before IPO. Annualised revenue reached $30 billion in April, up from $9 billion at end of 2025. The capital is following the operating momentum, and investor demand is described as overwhelming.

My PoV: OpenAI’s multi-cloud shift gives enterprise customers more negotiating leverage but also means the AI platform landscape is less settled than it appeared six months ago. Anthropic at $900 billion on $30 billion in revenue implies the market is pricing for infrastructure dominance, not software margins. If you are making multi-year platform commitments this year, you are making a bet on which infrastructure wins. Make it with eyes open.

2. The Physical Build Is Being Underestimated

Storage is the unglamorous backbone of AI. SanDisk posted Q3 revenue of $5.95 billion, up 97% year-on-year, beating estimates by over $1.2 billion. Its Datacenter segment more than tripled to $1.47 billion. Western Digital and Seagate told the same story: AI storage demand is running ahead of supply. Every model trained, every inference served, every dataset retained requires it. The component layer is generating returns the market had not priced.

India is building the next layer of global capacity. A Morgan Stanley report projected India’s data centre capacity will surge sixfold to 10.5 GW by FY2031, with AI workloads accounting for 6.8 GW. The capex pipeline is $60 billion. Data localisation policy and geopolitical realignment are accelerating what market demand alone would take longer to deliver. India is not treating data centre capacity as an afterthought. It is treating it as a prerequisite.

Meta acquired Assured Robot Intelligence, a startup building foundation models that enable robots to understand and adapt to human behaviour in dynamic environments. The team joins Meta Superintelligence Labs alongside Meta Robotics Studio. With Muse Spark already launched, $115 to $135 billion in 2026 capex, and now a robotics intelligence team inside its AI division, Meta’s direction is unambiguous: AI that acts in the physical world, not just the digital one.

My PoV: Storage, power, data centres, robotic intelligence. These are the layers that determine whether AI capability translates into AI deployment at scale. They are not glamorous, but they are where advantage is being built. Understanding your organisation’s dependency on each of these layers, including your supply chain’s exposure, is becoming as strategically important as choosing a model provider.

3. The Same AI, Two Edges

The UK government’s Cyber Security Breaches Survey 2025/26 found that 43% of British businesses, around 612,000 organisations, suffered a cyber breach or attack in the past year. Phishing dominates at 38%, and practitioners are explicit: AI tooling is making attacks more targeted, more personalised, and harder to detect. The breach rate has held flat for two years, which sounds stable. It is not. Methods are improving; defences are not. The incidents at M&S, Co-op, and Harrods cost an estimated £440 million combined. Only 15% of businesses review supplier risk. The tail is heavy.

The same AI is also reshaping the economics of building companies. Sam Altman observed that a new generation of startups is investing in compute rather than headcount, with founders in India attempting “zero person” startups where AI handles software, legal, and customer operations. AI agents are already trusted with multi-day knowledge work tasks. This compression will reach enterprise organisations with a lag, and when it does, the question will not be whether to restructure. It will be whether the operating model has been redesigned to compound the gains.

My PoV: AI lowers the cost of doing harm and the cost of doing work simultaneously. The organisations that navigate this well will treat cybersecurity and AI productivity as one strategic conversation, not two separate budget lines. The gap between AI-assisted attackers and AI-unaware defenders is not closing. Closing it is a leadership decision, not a technology one.

My Takeaway This Weekend

The AI investment cycle has crossed from promise into proof. The earnings are real. The infrastructure build is real. The consequences, in cyber risk, in platform restructuring, in operating model compression, are real and moving at the same pace as the capital.

For technology leaders, the question is no longer whether to engage. It is whether your vendor strategy, security posture, infrastructure roadmap, and operating model are calibrated for the speed at which this is arriving. April gave us the clearest signal yet that the window for unhurried decisions has closed.

Weekend Notebook #2617 – the Abbey Road Studios

amitabh apte

Published on LinkedIn and amitabhapte.com | 25th April 2026

This Weekend’s Notebook comes from a unique place. Abbey Road Studios in London, where The Beatles, Pink Floyd, Oasis and Adele recorded some of their most iconic music. It is a place where creative shifts became cultural shifts.

That felt fitting. Because the conversation this week was about a different kind of shift.

I participated in a panel titled “AI and the Evolving Landscape of Enterprise Risk” with Komal Mathur, Strategic Transformation Lead at SAP, and Gary Osborn, Head of Information Security at Amnesty International, moderated by well-known technology journalist, Mark Chillingworth. We discussed AI strategy, its implications for business organisations and its people, and the emerging governance landscape.

The panel was part of the two-day CIO event hosted by HotTopics, a well-regarded forum for CIOs, CTOs, CISOs, CDOs and technology leaders across the UK and increasingly Europe. The Studio event covered a wide range of themes: AI strategy and governance, cybersecurity and resilience, operating model transformation, talent and skills, and the growing role of technology leadership in shaping business outcomes.

I am sharing a few reflections from the panel and the broader event for the benefit of my network.

My Panel — From AI Tools to AI Systems

The core shift we discussed is simple, but not yet widely understood. AI is no longer just a tool. As it moves into core business systems; supply chain, finance, operations, it starts shaping decisions, not just supporting them. That changes the nature of risk.

For the past two years, most organisations have focused on data risk: leakage, privacy, compliance. Important, but incomplete. The next layer of risk is operational.

A chatbot giving a wrong answer is manageable. An agent taking a wrong action inside a live system is not. Especially at speed and scale.

We are moving from systems that inform decisions to systems that increasingly participate in them. That requires more than policies and principles. It requires governance built into the architecture itself: visibility, control points, and the ability to intervene when systems behave in unexpected ways.

Another theme that came through clearly was the tension leaders are navigating. Business wants speed and advantage. Technology teams want control and stability. Society expects accountability and trust. Balancing all three is becoming a core leadership challenge, and the CIO and the tech leaders are increasingly sitting at the intersections. 

The CIO role is quietly evolving into something closer to a Chief AI Business Risk Officer, whether organisations formalise that or not. Someone has to hold accountability for what autonomous systems do inside the enterprise. That needs to be a business leader with technical depth, not a policy document sitting in a shared drive.

Signals from the Room

Beyond the panel, a few patterns stood out across the two days.

The conversation around AI is maturing. The tone has shifted from “what can it do?” to “how do we manage it responsibly at scale?” Leaders are more candid now about the gap between ambition and readiness.

Cyber and resilience remain front of mind. As systems become more connected and more autonomous, the blast radius of failure increases. Security is no longer a layer. It is a design principle.

Governance is emerging as a differentiator, not a constraint. The ability to scale AI safely and predictably is becoming as important as the ability to deploy it. The organisations pulling ahead are the ones who figured that out first.

Operating models are under real pressure. AI is not just changing tools. It is reshaping workflows, decision rights, and team structures. Most organisations are still working this out, and the cultural change is proving harder than the technology.

And talent remains the binding constraint. Not just technical skills, but leaders who can connect technology, business and risk in the same conversation. That gap came up repeatedly. 

The sustainability aspects of technology, specifically the meaningful upcycling of hardware, devices, peripheral equipment, not just recycling is a cause which CIOs and tech leaders feel worthy of getting behind. The digital divide, the gap between people living at the edge of technology innovation and still a large population feeling left behind needs to be addressed as industry priority. 

The keynote from Eddie “The Eagle” Edwards, who finished last at the 1988 Winter Olympics with borrowed equipment and no support, was a reminder that showing up when nobody expects you to is its own form of leadership. His line stayed with the room: the people who said he did not belong were the same ones who wrote the rules about who was allowed to try.

My Takeaway This Weekend

The strongest signal across the two days was this: AI is moving from the edge of the organisation into its core systems. From answering questions to influencing decisions. Most organisations are still governed for the first. They will need to adapt quickly for the second.

AI risk is no longer about what the model says. It is about what the system does.

The leaders who will move ahead are not those deploying the most AI. They are the ones designing systems, operating models, and governance structures that can handle it responsibly.

Abbey Road Studios was a good reminder. The best work made in those studios was not made by avoiding risk. It was made by people who understood their craft well enough to take deliberate ones.

Weekend Notebook #2616 – When the Signal Meets the Noise

amitabh apte

Published on LinkedIn and amitabhapte.com on 19th Apr 2026

Three stories this week. One company managing simultaneous geopolitical pressure from three directions. A financial system confronting a risk it did not design for. And a market that is struggling to tell the difference between genuine AI transformation and an opportunistic rebrand. Each deserves a clear head.

1. Anthropic: Three Moves in One Week

Claude Opus 4.7 launched as Anthropic’s most capable generally available model, with step-change gains in agentic coding and complex engineering. Anthropic was explicit that Opus 4.7 trails the unreleased Mythos Preview. The release is also a governance experiment: the company deliberately reduced Opus 4.7’s offensive cyber capabilities during training and is using real-world deployment to test the guardrails it will eventually need for Mythos-class models at scale.

On the same day, Anthropic announced a 158,000 square foot London office for 800 staff, four times its current UK headcount, in the Knowledge Quarter alongside DeepMind, Meta, and OpenAI, which announced its own permanent London hub days earlier. The move deepens Anthropic’s work with the UK AI Security Institute, which evaluated Mythos Preview this week. With a Pentagon blacklisting in force, London is becoming both a talent hub and a political hedge.

The third move was a product one. Anthropic launched Claude Design, a new experimental tool that lets users create prototypes, slides, one-pagers, and visual assets through conversation. The target audience is explicitly non-designers: founders, product managers, analysts. You describe what you want, Claude produces an initial version, and you refine from there. It is a small product launch in the context of a week dominated by Mythos, but it is strategically coherent. Anthropic is quietly expanding its surface area from developer infrastructure into the everyday workflow of knowledge workers, the same ground occupied by Microsoft Copilot and Google Workspace AI.

My PoV: Anthropic is simultaneously managing a product too powerful to release publicly, a major geographic expansion, and a steady move into enterprise workflows. The Opus 4.7, London office, and Claude Design announcements are not separate stories. They are three layers of the same strategy: govern the frontier carefully, plant the flag in key markets, and expand the addressable use case before competitors consolidate their positions.

2. Finance Confronts a Risk It Didn’t Build For

At a G30 session on the sidelines of the IMF spring meetings, Barclays CEO CS Venkatakrishnan was direct: “On Mythos, it’s a serious issue. There will be a Mythos 2 and a Mythos 3, and they’ll come with probably distressing frequency.” His concern is the trajectory, not the model. Legacy banking infrastructure was not designed for an environment where AI can identify and chain software vulnerabilities at pace. The UK government simultaneously issued an open letter to businesses citing the AI Security Institute’s assessment that Mythos is “more capable at cyber offence than any model we have previously assessed.”

On a different but connected front, France’s Finance Minister called for more euro-denominated stablecoins and urged EU banks to accelerate tokenised deposits, naming the risk of “digital dollarization” directly. A 12-bank consortium including ING, UniCredit, and BNP Paribas is targeting a MiCA-compliant euro stablecoin in H2 2026. This is a policy reversal: France’s previous position was that private stablecoins had no place on European soil. Dollar-pegged tokens circulate at over $310 billion. Euro equivalents total under $1 billion. European policymakers have decided the cost of continued inaction now exceeds the risk.

My PoV: These two stories are more connected than they appear. Both reflect a financial system built for slow-moving threats and predictable regulatory cycles, now confronting neither. Defenders must succeed every time; attackers only once. For boards and risk committees, the right question is not whether Mythos itself is the threat. It is whether your security posture and your digital infrastructure were designed for the world Venkatakrishnan is describing.

3. Signal and Noise, Harder to Tell Apart

The tech market had a remarkable week. Oracle rose 27%, AMD climbed 42% over 13 consecutive sessions, and Microsoft posted its best week since 2015. Some of this reflected geopolitical peace hopes. But the underlying driver was real: Oracle’s cloud infrastructure revenue grew 84% year-on-year, AMD’s data centre GPU share is genuinely expanding, Azure is accelerating on AI workloads. The software sector is still down 19% year-to-date on AI disruption fears. The companies rebounding are the ones showing they know which side of that disruption they are on.

Then there is Allbirds. The wool sneaker brand, once valued at over $4 billion, announced it is selling its footwear assets and rebranding as “NewBird AI,” a GPU-as-a-service provider. It raised $50 million in convertible financing. Its stock rose 582% before falling 30% the next day. The company has no AI infrastructure expertise, no cloud customer base, and no obvious path to competing with CoreWeave or the hyperscalers. This is not an AI story. It is a distress story wearing an AI badge, and it follows a pattern that anyone who watched the dot-com era will recognise.

Tesla launched robotaxis in Dallas and Houston this weekend with unsupervised Model Y vehicles. The direction is right. The operational reality is modest: one active vehicle in each city at launch, four days before Q1 earnings. Waymo runs over 500,000 paid rides weekly across eleven cities. Tesla’s Austin fleet reached 46 vehicles after nearly a year and logged 14 crashes. The gap between the announcement and the business is still large.

My PoV: Oracle and AMD are rebounding on real infrastructure economics. Allbirds is the 1999 “.com rename” in 2026 clothing. Tesla’s robotaxi expansion is directionally meaningful but operationally early. The discipline of separating these is not a nice-to-have. The market is currently rewarding the story and the substance equally, and that gap will close.

My Takeaway This Weekend

AI capability is advancing faster than governance can absorb. The infrastructure sustaining it is generating real returns. And the speculative energy around the narrative is producing decisions that deserve more scepticism than they are getting. All three are true simultaneously. The leaders who navigate this well are the ones who can hold all three frames at once, without collapsing into either uncritical enthusiasm or reflexive caution.

Weekend Notebook #2615 – When AI become Infrastructure, Risk and Rivalry

amitabh apte

Published on LinkedIn and amitabhapte.com on 12th Apr 2026

This was a week where AI showed up as an infrastructure bet, systemic risk, competitive battleground, and talent story, all at once. Many stories. One consistent thread: the foundational layers of the AI economy are being built and contested simultaneously, and the institutions designed for a slower world are catching up in real time.

The Anthropic Week: Revenue, Risk, and Rivalry

Three distinct signals from one company. First, the commercial: Anthropic’s annualised revenue crossed $30 billion, up from $9 billion just four months ago. CoreWeave sealed a multi-year infrastructure deal to power Claude workloads, days after a $21 billion commitment from Meta. Nine of the ten leading AI model providers now run on CoreWeave’s platform. Infrastructure is consolidating fast.

Second, the risk signal. Anthropic introduced Mythos Preview, a model so capable at finding and exploiting software vulnerabilities that the company chose not to release it publicly. Under Project Glasswing, access is limited to Amazon, Apple, Google, Microsoft, JPMorgan, and around 40 critical infrastructure organisations. The model has already identified vulnerabilities across every major operating system and browser, including a 27-year-old flaw in OpenBSD. Treasury officials and the Federal Reserve convened an emergency meeting with Wall Street’s senior bank CEOs. The Bank of England placed Mythos on the agenda of its Cross-Market Operational Resilience Group, alongside the FCA and the National Cyber Security Centre. Canada convened its own session the same week.

Third, the investor story. OpenAI’s secondary market shares have become difficult to sell. Around $600 million of stock found very few buyers on the secondary market. Meanwhile, demand for Anthropic shares is described as almost insatiable, with $2 billion in declared buy interest and almost no sellers. OpenAI responded with an investor memo characterising Anthropic as compute-constrained. The defensiveness itself is the signal.

My PoV: Mythos is the clearest signal yet that AI safety is an operational risk category, not a philosophical one. For technology leaders, the question is not whether your organisation uses Anthropic products. It is whether your security posture has been updated for an era where AI can identify and weaponise software vulnerabilities at machine speed. On the investor story: the AI platform you build on today is not easily changed. Governance clarity and consistent product performance now matter as much as benchmark scores.

New Entrants, New Approvals

Meta’s $14.3 billion bet on Alexandr Wang delivered its first output this week. Muse Spark, the first model from Meta Superintelligence Labs, is a natively multimodal reasoning model rebuilt from the ground up over nine months. It is competitive with frontier models on several benchmarks, though not a leader across the board. More significant than the model is the strategy: Muse Spark launched as a closed, proprietary product. Meta, which built its AI identity on open-source Llama, has quietly changed its approach. With capital expenditure planned at $115 to $135 billion in 2026, nearly double last year, and three billion daily users as a distribution surface, Meta is no longer treating AI as an experiment.

Separately, the Netherlands became the first EU country to formally approve Tesla’s Full Self-Driving Supervised system, after 18 months of testing covering 1.6 million kilometres on European roads. The system is not autonomous: the driver remains legally responsible and must be ready to intervene. But the approval, under EU mutual recognition rules, opens a pathway to continent-wide rollout by mid-2026. It is the first time a physical AI system of this complexity has passed rigorous European regulatory scrutiny, and the precedent will matter well beyond vehicles.

My PoV: Meta’s shift from open to closed signals that distribution advantage, not model openness, is where the competitive moat is now being built. For enterprise leaders, the Tesla approval matters less as a driving story and more as a governance template. Physical AI systems require documented safety evidence, long evaluation windows, and ongoing reporting obligations. Build that infrastructure now, before the regulator requests it.

India’s Technical Capital Comes of Age

Two data points deserve to be read together. GitHub reported that India now has 27 million developers on its platform, 15 percent of the global total, with more than two million new joiners in 2026 alone, more than any other country. India is the world’s second largest contributor to open-source AI projects, with over 7.5 million contributions on GitHub. At the same time, TCS posted Q4 results showing $12 billion in contract value for the quarter, $40.7 billion for the year, and annualised AI revenue of $2.3 billion. Its HyperVault data centre business, targeting 1 gigawatt of capacity, has moved into commercial structuring with hyperscalers and frontier AI companies. The positioning is explicit: infrastructure to intelligence, end to end.

My PoV: India is simultaneously facing the erosion of traditional IT outsourcing as AI automates entry-level tasks, and building the technical and infrastructure base to compete in the next generation of AI deployment. A country producing 27 million GitHub developers and the world’s second largest open-source AI contributor base is not a back office. It is a source of technical capital at a scale few geographies can match. Enterprise talent strategies that are not designed to work with that pipeline are working around it at significant cost.

My Takeaway This Weekend

The model layer of AI is commoditising quickly. The infrastructure layer, physical, computational, regulatory, and human, is not. The companies and countries securing advantaged positions in those foundational layers will shape the AI decade. The ones still treating AI as a product decision will find themselves working within a landscape that others have already built.

The Mythos story, the CoreWeave deals, the Tesla approval, India’s developer numbers, Meta’s infrastructure bet: none are separate stories. They are all evidence of the same transition. Intelligence is no longer arriving as a feature. It is arriving as a structural condition. The leadership question is no longer whether to engage. It is whether your organisation is building on the right foundations before the terrain gets harder to move on.

Weekend Notebook #2614 – when AI sets the Terms

amitabh apte

Published on LinkedIn and amitabhapte.com  on 5th April 2026

Three forces are running simultaneously through the AI landscape right now: a scramble for physical control of the underlying infrastructure, a repricing of how AI-driven value gets negotiated, and a fracture in how the world governs it. This week’s signals belong to all three.

1. The Race to Own the Substrate

New IDC data reviewed by Reuters shows Chinese GPU and AI chip makers captured 41% of China’s AI accelerator market in 2025. Huawei alone shipped around 812,000 chips, with Alibaba’s T-Head and Baidu’s Kunlunxin growing behind it. Nvidia still leads at 55%, but the retreat is real. 

The same instinct is driving infrastructure investment across Asia. Airtel raised $1 billion for its data centre arm Nxtra from Carlyle, Alpha Wave, and Anchorage Capital, targeting a scale-up to 1GW of capacity in India, a country that already has Google’s $15 billion data centre commitment and a 20-year tax holiday for hyperscalers. Microsoft’s $10 billion commitment to Japan through 2029 is structured similarly: AI infrastructure, national cybersecurity cooperation, and data processed inside Japan’s borders. These are sovereignty arrangements as much as commercial deals. 

2. AI Is Changing the Terms of Access

In enterprise software, ServiceNow’s CEO Bill McDermott has repositioned the company around a single argument: AI models identify problems but struggle to execute reliably across governed, auditable enterprise workflows. ServiceNow owns the last-mile execution layer, and its Now Assist product is tracking toward a $1 billion annual run rate. The model shift underneath it is structural: from per-seat licensing to outcome-based pricing, what McDermott calls digital labour. 

In capital markets, the dynamic is rawer. SpaceX is targeting a valuation above $2 trillion in what could be the largest IPO in history. According to the New York Times, Musk has required the lead banks, Morgan Stanley, Goldman Sachs, JPMorgan, Bank of America, and Citigroup, to purchase Grok subscriptions as a condition of the mandate. Some have agreed to spend tens of millions per year. AI adoption bundled into the price of deal access is a new distribution model. 

3. AI Governance Is Becoming a Procurement Reality

Several jurisdictions are now introducing procurement standards that require AI vendors to demonstrate safety and privacy safeguards before accessing public sector contracts. California’s Executive Order N-5-26 is a recent example: companies seeking state contracts must disclose safeguards against harmful content, bias, and civil rights violations, with agencies conducting their own independent assessments. The practical result is a growing patchwork of standards that enterprise technology teams will have to navigate across geographies. California’s procurement rules have a history of becoming de facto global benchmarks, much as GDPR did from a single jurisdiction. Any organisation selling AI products into regulated markets or the public sector should treat this as an active compliance question, not a future one.

4. Japan: Automation as Survival, Not Disruption

TechCrunch reported this week on Japan’s push into physical AI, driven by demographic emergency rather than efficiency ambition. With over 28% of the population above 65 and a working-age population contracting annually, Japan is deploying robots to fill positions that cannot be staffed. The government has committed $6.3 billion and is targeting 30% of the global physical AI market by 2040. The industry signal is unambiguous: customer-paid deployments, full-shift operation, measurable productivity. Japan is the clearest preview of what automation looks like when labour scarcity, not cost reduction, is the driver. 

My Takeaway This Weekend

The infrastructure layer of AI is being claimed physically and politically. The business models above it are repricing around outcomes. The governance frameworks meant to contain it are fracturing. And the labour markets it will reshape are already showing, in Japan, what comes next. AI strategy, infrastructure strategy, talent strategy, and governance strategy are now the same conversation. Running them separately is how organisations fall behind without noticing.

Weekend Notebook #2613 – When AI Meets Friction

amitabh apte

Published on LinkedIn and amitabhapte.com on 29th Mar 2026

This week, ambition collided with reality across nearly every front of the AI story. Agentic commerce promised to remove the human from the checkout. A new model leaked before it was ready to launch. The internet crossed a threshold most of us hadn’t noticed. And one logistics giant did something quietly radical: it decided to teach half a million people to work with AI rather than step aside for it. Different signals. Same underlying tension. The gap between what AI can do and what organisations are actually ready for is growing. That gap is where leadership happens.

Agentic Commerce: The First Honest Report Card

Late last year, OpenAI launched Instant Checkout, a feature that lets shoppers complete purchases directly inside ChatGPT without ever visiting a retailer’s website. Walmart signed up as the launch partner. Etsy and Shopify quickly followed. The narrative was compelling: conversational commerce had arrived. The results were not. Walmart has now disclosed that conversion rates inside ChatGPT were three times lower than for click-out experiences that redirected users to Walmart’s own site. That is not a rounding error. It is a structural finding. OpenAI has since moved on, phasing out Instant Checkout in favour of an app-based model that gives retailers more control of the transaction. Walmart is now embedding its own chatbot, Sparky, directly into ChatGPT and Google Gemini, rather than handing the checkout process to a third party.

Meanwhile, Gap has become the first major fashion retailer to launch direct checkout within Google’s Gemini platform, part of an emerging Universal Commerce Protocol that Google has been rolling out since January. The approach is different in intent: Gap is pairing the checkout integration with an AI-powered sizing tool, specifically targeting the return rate problem that plagues online apparel. Net sales were up 2% in Q4 2025, with online sales growing 5%. The CTO was clear that this is about solving real customer problems, not chasing innovation for its own sake.

My PoV: The Walmart data is the most useful signal the agentic commerce story has produced. Consumers will use AI for discovery. They are not yet ready to surrender the checkout to it. The friction of a familiar interface, seeing the full cart, entering payment details on a trusted site, provides reassurance that an embedded AI flow does not yet replicate. For enterprise and retail leaders, the lesson is architectural: own the transaction layer. Let the AI own the discovery. The retailers now embedding their own branded experiences into AI platforms, rather than ceding the whole journey, are making the smarter structural bet.

Anthropic’s Week: Values as a Growth Strategy

It has been a remarkable few months for Anthropic. A public standoff with the US Department of Defense over the use of Claude in lethal autonomous systems, followed by Super Bowl ads that went after OpenAI’s decision to serve ads to its users, has produced something unexpected: a subscriber surge. TechCrunch analysis of 28 million US consumer transactions shows paid subscriptions more than doubling since the start of 2026, with record new sign-ups in January and February. Web traffic was up 43% month-on-month in February and nearly tripled year-on-year. Most new subscribers are at the entry-level Pro tier at $20 per month. Claude Code and Claude Cowork, the developer and productivity tools released in January, have accelerated that growth further.

The same week brought a different kind of Anthropic headline. The company inadvertently exposed an internal draft blog post in a publicly searchable data store, revealing a new model under development called Claude Mythos. The draft described it as the company’s most powerful model to date, part of a new capability tier called Capybara, significantly beyond the current Opus tier. The document also described the model as posing unprecedented cybersecurity risks, specifically for its ability to identify and exploit software vulnerabilities at speed. Cybersecurity stocks fell immediately: CrowdStrike, Palo Alto Networks and Zscaler each dropped around 6%. Anthropic confirmed the model exists and is being tested with early access customers.

And underpinning all of this, a new report from Human Security found that AI and automated traffic have, for the first time, overtaken human traffic on the internet. Automated traffic grew eight times faster than human traffic in 2025. AI-driven traffic alone grew 187% across the calendar year. The internet was built on the assumption that a human being was on the other side of the screen. That assumption is no longer safe.

My PoV: Three separate but connected signals from Anthropic this week. First, values can be a growth driver. Taking a principled public position on how AI should and should not be used attracted consumers in a way that a model benchmark never could. That is worth paying attention to for any enterprise working out how to position itself in the AI market. Second, the Mythos leak reminds us that the cybersecurity stakes are rising with every capability jump. Anthropic’s plan to give cyber defenders early access before general release is the right instinct, but the gap between what AI can do and what defenders are prepared for is narrowing fast. Third, if the majority of internet traffic is now non-human, the infrastructure assumptions of most enterprise digital strategies need revisiting, from fraud detection to API design to web analytics.

AI and the Financial System: A Stress Test No One Planned For

While AI companies attract record investment, a less discussed story is developing in private credit. Shadow banking, the network of private credit funds, business development companies and non-bank lenders that has grown significantly since 2008, has been heavily exposed to software-sector loans. The concern, now surfacing in the mainstream, is that AI may be systematically undermining the value of the software companies these funds have lent against. Apollo Global Management was among the first to flag it publicly last year: “Is software dead?” is now the question private credit managers are trying to price. A closely watched index of 44 business development companies shed around $5 billion in February. The Bank of England has announced it will conduct the world’s first stress test of the shadow banking sector. Lloyd Blankfein has drawn parallels to 2005 and 2006, when hidden leverage was building quietly beneath a rising tide.

My PoV: This story rarely appears in AI newsletters. It should. The thesis is straightforward: if AI can replace significant portions of software development work, the revenue and margins of many mid-market software companies, which form the collateral base for billions in private credit, come under structural pressure. This is not a prediction of an imminent crisis. It is an observation that the financial system has not yet priced AI disruption into the sectors most exposed to it. For enterprise technology leaders, this has a practical implication: the cost, availability and terms of technology financing are likely to become more volatile. AI is not just disrupting products. It is beginning to reprice the capital structures behind them.

The Workforce Question: FedEx Bets on Learning

Against a backdrop of sector layoffs and automation anxiety, FedEx has launched what may be the largest corporate AI upskilling programme in logistics. The initiative covers more than 400,000 employees globally, with personalised, role-based modules that will update monthly. The programme is explicitly tied to internal promotion pathways. The company calls it “promotion-ready” AI training. Frontline workers are already applying for corporate roles at higher rates since the programme launched. Every C-suite executive at FedEx spent two days in Silicon Valley selecting the right technology partners before a single module was deployed.

This matters in context. UPS announced 30,000 layoffs on top of 48,000 the previous year. FedEx has also made cuts. But its strategic posture is distinctly different: use AI to make the workforce more capable, not smaller. The company measures something it calls AIQ, an AI quotient, tracking progress rather than just completion rates. Chief Data and Information Officer Vishal Talwar was direct: “We are measuring progress around AI, not necessarily just success, because it’s going to be very difficult to say this success is only attributed to AI.”

My PoV: Only 28% of organisations have embedded continuous AI learning, according to Accenture’s 2026 Pulse of Change report. FedEx is in that minority, and it is moving at a scale that few others have attempted. The principle behind the programme is the right one: AI literacy cannot be a specialist skill. It needs to be a baseline capability across every level of the organisation, from warehouse floor to boardroom. The harder question is how to measure the business value generated rather than just the learning hours logged. That measurement challenge is the next frontier of enterprise AI investment discipline.

My Takeaway This Weekend

Four stories, one shared theme: AI is not meeting reality gently. Agentic commerce stumbled at the transaction layer, where trust has always mattered most. A powerful new AI model leaked before its makers were ready, and the market reacted to the risk before the product even shipped. The internet crossed a threshold that few enterprise strategies were built for. And in logistics, one company decided to bet on its people rather than against them.

The leaders navigating this well are not those with the most ambitious AI roadmaps. They are the ones who are honest about where friction is real, where trust has not yet been earned, and where their own organisations need to build capability before deploying it. Friction is not a failure of technology. It is the system telling you where the work still needs to be done.

Weekend Notebook #2612 – when Agents become the Architecture

amitabh apte

GTC Live 2026 Keynote Pregame – photo credit NVIDIA GTC

Published on LinkedIn and amitabhapte.com on 22nd Mar 2026

GTC 2026 drew 30,000 people to San Jose. Jensen Huang announced $1 trillion in confirmed orders for Blackwell and Vera Rubin systems through 2027, double last year’s projection. But the number was not the headline. The architecture behind it was.

From Data Centres to AI Factories

Huang reframed the data centre entirely. The new construct is the AI factory, a facility whose primary output is not storage or compute, but tokens. Every query answered, every decision supported, every automated workflow consumes them. The new efficiency metric is not uptime. It is token throughput per watt.

This changes the business case for infrastructure investment. Data centres were cost centres. AI factories are production lines. When the output has a unit price, the conversation with the business shifts fundamentally.

Nvidia’s keynote slides showed 40% of its order pipeline now coming from enterprise, sovereign AI, and industrial customers, not just hyperscalers. The enterprise wave is no longer coming. It has arrived.

My PoV: CIOs who still frame infrastructure purely as a cost management conversation are using the wrong model. Token economics and inference costs belong in your architecture discussions now. Your business leaders will ask about them within 18 months.

The Agent is a Platform

The most important slide of the keynote carried a simple title: ‘Agents: A New Computing Platform.’ Huang’s argument was precise. The PC was a platform. The smartphone was a platform. The agent is next, with its own architecture: a reasoning core connected to memory, sub-agents, tools, files, and a multi-modal prompt layer.

Nvidia made this concrete with NemoClaw, an enterprise-ready implementation of the OpenClaw agentic framework, bringing autonomous agents inside the enterprise firewall with privacy controls and policy guardrails. Huang also noted that Nvidia’s own engineers will receive annual token budgets as a productivity metric. Token consumption is becoming a measure of knowledge work output.

My PoV: If the agent is a platform, enterprise architecture must be designed around it, not retrofitted for it. The question is not which agent tool to pilot. It is what your data, security, and integration architecture looks like when agents become the primary consumers of enterprise systems.

Enterprise IT: From SaaS to Agent-as-a-Service

One slide showed two pictures. On the left: today. Data centre, SaaS software, GSI, humans on top. On the right: tomorrow. An AI factory generating tokens, software and AI providers connected by agents, and humans repositioned as enterprise information workers directing and overseeing rather than executing.

Huang called this the Enterprise IT Renaissance. Not disruption. Not replacement. Renaissance. And the implications for software were explicit: every SaaS company must become an Agentic-as-a-Service company. The subscription model built on human users logging in is giving way to a consumption model built on agents accessing capabilities programmatically.

My PoV: This is the most significant vendor landscape shift since the move to cloud. The roadmap conversations you have with your major software vendors over the next twelve months should be explicitly about their agentic strategy. If they do not have a credible one, that is a signal worth taking seriously.

My Takeaway This Weekend

GTC 2026 was not a product launch. It was an architectural declaration. The era of AI as a layer on top of existing systems is ending. What follows is AI as the foundation, with tokens as the unit of value, agents as the primary computing paradigm, and enterprise IT reborn around AI factories.

The organisations that rearchitect early will compound advantage. Those that treat this as another cycle to manage carefully will find the gap harder to close each quarter.