Tag: ai

Weekend Notebook #31 – AI’s Hard Power: Data Centres, Defence, and Design

amitabh apte

Published on LinkedIn and amitabhapte.com on Sunday, 3rd August, 2025

In spotlight this week: The age of infrastructure – AI’s physical footprint

This quarter, Big Tech’s capital spending on AI infrastructure reached historic levels. Meta, Microsoft, Amazon, and Google collectively spent nearly $100 billion on data centres, chips, and hardware, more than consumer spending contributed to GDP growth. OpenAI’s Stargate Norway project, housing 100,000 Nvidia GPUs, exemplifies this shift. See quarterly earnings highlights in the later parts of this article.

What’s Happening? We’re seeing a dramatic shift in how AI is being built and scaled. The focus is no longer just on algorithms or model performance, it’s on physical infrastructure. Data centres, energy grids, GPU clusters, and sovereign compute zones are becoming the new battlegrounds. OpenAI’s Stargate project is emblematic of this shift: a hyperscale facility designed to power frontier models with industrial-grade reliability.

Why does this matter? This is the moment AI becomes tangible. It’s not just software, it’s steel, silicon, and electricity. The implications are vast:

  • Economic: AI infrastructure spend is now a macroeconomic force, influencing GDP and reshaping capital markets.
  • Geopolitical: Countries are racing to secure compute sovereignty, energy access, and chip supply chains.
  • Enterprise: For business and technology leaders alike, infrastructure strategy is now core to AI strategy. It’s no longer just about cloud contracts, it’s about latency, throughput, and deployment architecture.

My point of view – We’re entering a new industrial era, one where compute is the new oil, and data centres are the new ports. This isn’t just about digital transformation; it’s about physical transformation. For business and technology leaders, this means thinking beyond models and prompts. It’s about power, land, logistics, and latency. The infrastructure layer is where the next competitive moats will be built.

Noteworthy this week: what caught my eye in AI and tech world

OpenAI’s $8.3B raise and valuation – OpenAI has raised $8.3 billion at a $300 billion valuation, with annual recurring revenue now at $13 billion. The Stargate data centre network is expanding into Europe, with Norway chosen for its hydropower and low energy demand. My PoV: This shift of OpenAI from API access to full-stack infrastructure is redefining what it means to be an AI company. And it’s a reminder that the winners in this space will be those who can scale both intelligence and infrastructure.

Microsoft hits $4T milestone – Microsoft’s stock surged past the $4 trillion mark following strong earnings, joining Nvidia in an exclusive club. Azure revenue topped $75 billion, up 34% YoY, and the company posted its fastest growth in over three years. My PoV: This is a milestone not just for Microsoft, but for enterprise AI. The company’s ability to integrate AI across its stack, from Copilot to Azure to GitHub, is translating into real revenue and market dominance. It’s also a signal that the GenAI wave is no longer hype, it’s hitting the balance sheet.

Figma’s explosive IPO – Figma’s IPO stunned Wall Street, with shares soaring 250% on debut and closing at a valuation near $60 billion. It’s the biggest design software IPO in history, and a comeback story after Adobe’s failed $20B acquisition in 2023. My POV: Figma’s success shows that design is no longer a niche—it’s infrastructure for the digital economy. In a world of AI-generated content, the tools that shape experience and interface are more valuable than ever. This IPO also signals a thaw in the tech IPO market, with design leading the charge.

Palantir’s Army contract – Palantir secured a $10 billion contract with the U.S. Army to consolidate 75 separate deals into one enterprise framework for software and data needs. My POV: This is defence AI at scale. The deal reflects how AI is becoming foundational to national security, and how enterprise platforms are being reimagined for battlefield intelligence. It’s also a reminder that the AI race isn’t just commercial, it’s geopolitical.

Earnings Pulse: Microsoft, Apple, Meta, Amazon

Microsoft Q2 earnings – Cloud and AI drove a blockbuster quarter. Azure revenue hit $75B, and the company returned $9.7B to shareholders.

Apple Q2 earnings – Posted $95.4B in revenue, up 5% YoY, with record services growth and strong iPhone 16e sales. But China softness and tariff concerns linger.

Meta Q2 earnings – Revenue jumped 22% to $47.5B, with strong ad growth and a 36% rise in net income. Zuckerberg teased “personal superintelligence” as the next frontier.

Amazon Q2 earnings – Delivered $167.7B in revenue, up 13%. AWS grew 17.5%, and CEO Andy Jassy spotlighted new AI agents like Kiro and Strands as key to future growth.

My POV: The earnings season confirms it: AI is now a revenue engine, not a research project. But the divergence is clear; Microsoft and Meta are pulling ahead on infrastructure and monetisation, while Apple and Amazon are still translating AI into product and platform wins.

In summary – my key takeaway this weekend

“The winners in AI won’t just scale intelligence—they’ll scale deployment.
It’s no longer about building smarter models; it’s about embedding them into infrastructure, products, and institutions. From hyperscale data centres to battlefield software and consumer platforms, AI is becoming the operating system of everything.”

Weekend Notebook #30 – Agents, Robotaxis, Windsurf, Scaling AI

amitabh apte

In Spotlight this Week: ChatGPT Agents-The Next Leap in Autonomous AI

This week, OpenAI introduced a significant upgrade inside ChatGPT: agents. These aren’t just smarter chatbots, they’re autonomous digital co-workers that can take action, not just provide answers.

So what are ChatGPT agents? Imagine assigning a task like “find the best flights under $800 and book one,” and the agent goes off to browse, fill out forms, download files, generate spreadsheets, or run code, all independently, securely, and within defined guardrails. It’s a major step beyond prompt and response.

Why does this matter? Until now, most AI systems have been reactive, you ask, it replies. With agents, we step into the realm of proactive AI. Tools that can reason, navigate real-world systems, and deliver outcomes. It’s not just an upgrade, it’s a rethink of how digital work gets done.

For digital and business leaders, this opens up new possibilities:

  • Deploying agents across finance, HR, marketing, or data ops
  • Freeing teams to focus on higher-order tasks like judgement, design, and decision-making
  • Building modular workflows that connect apps, documents, and tools without traditional integrations or code

Are Agents different that Agentic AI? – There’s an important distinction here. “Agentic AI” is the design philosophy, AI that plans, decides, and acts to achieve goals. What OpenAI has now launched is a concrete implementation of that vision. These agents live inside ChatGPT, wired into tools, memory, APIs, and your workspace. This is no longer theory. It’s operational.

This evolution will reshape how we approach AI in the enterprise. It changes how we think about roles, delegation, and execution. We’ll soon be designing teams where agents carry out tasks just like apps once did, only now, with autonomy and context.

Noteworthy this week: important developments across the AI and tech landscape

OpenAI has launched a $10M+ AI consulting business, embedding engineering teams inside enterprises to accelerate custom AI deployments. It marks a shift from simply offering access to models, toward driving hands-on business transformation. OpenAI isn’t just a tech vendor anymore, it’s aiming to become a full-stack AI delivery partner.

Google paid $2.4B in licensing fees to Windsurf, an AI coding startup, while simultaneously hiring away its top talent, including the CEO. The company remains technically independent, but gutted of its core team. It’s a striking example of how Big Tech is buying talent and capability without formal acquisitions. Another startup, Cognition, picked up the remainder of the team. Urgency in the AI arms race is clearly reshaping how innovation is scaled, and acquired.

Uber is investing more than $500 million in Lucid and Nuro to deploy a fleet of 20,000 AI-powered robotaxis over the next six years. It’s their biggest move yet toward owning autonomous mobility infrastructure and integrating AI into core transport systems, rather than relying on external platforms.

Meta appointed Shengjia Zhao, co-creator of ChatGPT and former OpenAI scientist, as chief scientist of its new Meta Super-intelligence Labs. Zhao will lead foundational AI research and long-term scaling. It signals Meta’s aggressive ambition to compete at the frontier of AI, with plans to invest hundreds of billions in compute and infrastructure.

Meanwhile, news publishers are facing major disruption from Google’s AI Overviews, which summarise information above traditional search links. Studies show this has led to a 79% drop in traffic for many media outlets. There’s growing concern that the economics of independent journalism may not survive in an AI-first search experience. It’s a reminder that even technically brilliant innovations need to be matched with models that protect context, attribution, and quality.

As always, the real challenge isn’t what the tech can do, it’s what we choose to do with tech.