November 7, 2025

This Week in AI (Nov 1–7, 2025): Why $360 Billion in AI Spending Means Fewer Jobs, Not More

Your Weekly Recap In A Nutshell

By Jay Cadmus

Infrastructure and Compute Become Strategic Priorities

OpenAI's $38 Billion AWS Commitment Signals Multi-Cloud Future

OpenAI signed a seven-year, $38 billion deal with Amazon Web Services to secure access to hundreds of thousands of Nvidia GPUs and EC2 UltraServers, with capacity scaling through 2026–2027 (Reuters, CNBC, Wired: 11/3/2025). The agreement marks a significant strategic shift for OpenAI, which has historically relied heavily on Microsoft Azure for its computing infrastructure. CEO Sam Altman emphasized that "scaling frontier AI requires massive, reliable compute," while AWS CEO Matt Garman stated the infrastructure would serve as "a backbone" for OpenAI's AI ambitions (CNBC: 11/3/2025).

The deal positions OpenAI to operate across multiple cloud providers rather than depending on a single vendor, a strategy that reduces both technical risk and pricing pressure. For enterprises, this signals an industry-wide shift toward multi-cloud approaches for AI workloads, particularly as GPU scarcity continues and hyperscalers compete for dominance in the AI infrastructure race (Reuters: 11/4/2025).

What makes this particularly significant is the timing and scale. As I detailed in The $7.8 Trillion Bet, hyperscalers are collectively committing over $360 billion annually to AI infrastructure in 2025, a 47% increase from 2024. OpenAI's $38 billion commitment to AWS is part of this broader mobilization, demonstrating that compute access has become a strategic differentiator. Companies are not building for today's capabilities; they're securing infrastructure for what becomes possible in 2027, 2030, and beyond (McKinsey, AIX Energy: 2025).

Amazon's stock closed at a record high following the announcement, reflecting investor confidence in AWS's positioning as a critical enabler of next-generation AI (CNBC: 11/3/2025).

For businesses evaluating their own AI infrastructure strategies, the takeaway is clear: compute access is becoming a competitive necessity. Companies should consider capacity reservations, multi-cloud partnerships, or long-term agreements to avoid being priced out or bottlenecked as demand continues to surge.

VAST Data and CoreWeave Sign $1.17 Billion Data Platform Deal

AI storage platform vendor VAST Data inked a $1.17 billion, multi-year agreement with cloud infrastructure provider CoreWeave, positioning VAST as CoreWeave's primary data foundation for AI cloud workloads (Reuters, Blocks and Files: 11/6/2025). The partnership integrates file, object, and database services into a unified data plane optimized for both training and inference at massive scale.

CoreWeave, which serves clients including Meta, OpenAI, and Microsoft, will leverage VAST's architecture to manage hundreds of petabytes of data with improved throughput and reduced bottlenecks (Reuters: 11/6/2025). The deal underscores an important shift in the AI infrastructure stack: data platforms are becoming just as critical as GPU access itself.

For organizations training large models or running high-volume inference workloads, the integration of compute and storage into tightly coupled systems is becoming the norm. Businesses evaluating GPU cloud options should benchmark not just compute performance but also data pipeline efficiency, as storage bottlenecks can significantly undermine the value of even the most powerful accelerators (Blocks and Files: 11/5/2025).

VAST's deal with CoreWeave also reflects the broader trend of vertical integration in AI infrastructure, where cloud providers are partnering with specialized vendors to deliver end-to-end solutions rather than expecting customers to stitch together their own stacks from disparate components.

Google and Anthropic Expand TPU Partnership Worth Tens of Billions

Anthropic announced plans to expand its use of Google Cloud's Tensor Processing Units (TPUs), with access to up to one million chips valued at tens of billions of dollars (VentureBeat, Economic Times: 11/5–6/2025). The deal will bring over a gigawatt of TPU capacity online in 2026, representing one of the largest AI infrastructure commitments to date.

Anthropic selected Google's TPUs for their price-performance advantages and energy efficiency, deepening a partnership that already includes Google's $3 billion investment in the AI safety startup (Reuters: 10/23/2025). Google's seventh-generation Ironwood TPU, which delivers more than 4x the performance of its predecessor, will be made available for public use in the coming weeks (VentureBeat: 11/6/2025).

This agreement positions Google's custom silicon as a viable alternative to Nvidia GPUs, potentially reshaping competitive dynamics in AI hardware. For Anthropic, it ensures the computing resources needed to train and serve next-generation Claude models to meet rapidly growing enterprise demand (Economic Times: 11/5/2025).

The broader implication is that vertical integration (from chip design through cloud services to AI models) is becoming the preferred strategy among hyperscalers. Companies relying exclusively on Nvidia hardware should evaluate diversification strategies, as supply constraints and geopolitical risks continue to create volatility in chip access.

Policy and Regulation Face Pressure and Uncertainty

EU Weighs Softening AI Act Under U.S. and Big Tech Pressure

The European Commission is considering a one-year grace period for some AI Act obligations and delaying transparency-related fines until August 2027 as part of a "simplification package" to be discussed November 19 (Financial Times, Fortune: 11/7/2025). The proposed adjustments respond to mounting pressure from U.S. authorities and major technology companies arguing that strict regulations could stifle innovation and hurt European competitiveness.

Under the draft proposal, companies deploying high-risk AI systems could receive additional time to make adjustments "without disrupting the market," while postponing enforcement would give firms more runway to adapt to transparency requirements (Fortune: 11/7/2025). The Trump administration has warned of potential retaliatory measures if the EU's rules are perceived as adversarial to U.S. interests, prompting informal discussions to align the AI Act with American trade priorities (Fortune: 11/7/2025).

Tech giants including Meta and Alphabet have criticized the Act's broad definitions of "high-risk" AI, warning that complex compliance requirements discourage experimentation and disadvantage smaller developers. European Commission spokesperson Thomas Regnier emphasized that "the Commission will always remain fully behind the AI Act and its objectives," but acknowledged that "a reflection is still ongoing" regarding targeted delays (Fortune: 11/7/2025).

For businesses operating in Europe, this creates both opportunity and uncertainty. A grace period provides breathing room, but long-term compliance obligations remain. Companies should continue building transparency documentation, model cards, and risk assessment frameworks even if near-term enforcement is delayed. The broader lesson is that regulatory fragmentation between the U.S., EU, and China will persist, complicating global operations and requiring region-specific compliance strategies (Wall Street Journal, Reuters: 11/7/2025).

Nvidia Confirms No Blackwell Chips for China

Nvidia CEO Jensen Huang confirmed there are "no active discussions" to sell the company's advanced Blackwell AI chips to China, stating "currently, we are not planning to ship anything to China" (Reuters, Bloomberg: 11/6–7/2025). The statement, made during Huang's visit to Taiwan, clarifies Nvidia's position amid ongoing U.S. export controls designed to limit China's access to frontier AI hardware for national security reasons.

While the U.S. has allowed Nvidia to sell its less-advanced H20 chip in China, Huang noted that Chinese authorities "made it clear they don't want Nvidia to be there right now," resulting in zero market share in China's advanced AI chip segment (Bloomberg: 11/7/2025). President Trump has taken a hardline stance, stating that Blackwell chips (which he described as "10 years ahead of every chip") would not be available to "other people" (Reuters: 11/2/2025).

The export restrictions are accelerating China's push to develop domestic alternatives. Chinese companies like Huawei are scaling massive chip clusters and leveraging low-cost energy to power AI independence, with plans to invest $70 billion in AI data centers and issue further guidelines integrating AI into manufacturing (CNBC, Goldman Sachs: 11/2–6/2025).

For enterprises dependent on frontier GPU access, this creates both challenges and imperatives. Supply tightness will persist globally as China is effectively locked out of Nvidia's cutting-edge chips. Companies should evaluate alternative accelerators from AMD and Intel, plan for hybrid training and inference topologies, and consider multi-vendor strategies to mitigate supply risk (Reuters: 11/6/2025).

The chip standoff reflects a broader U.S.-China technology rivalry shaping not just hardware access but also data policy, export controls, and global AI development trajectories.

Agentic AI and the Governance Crisis

Amazon Sues Perplexity Over Autonomous Shopping Agents

Amazon filed suit against Perplexity AI in U.S. District Court for Northern California, alleging the startup's Comet browser and AI agent unlawfully accessed customer accounts, masked automated activity as human behavior, and violated Amazon's terms of service (Reuters, PC Mag: 11/4/2025). Amazon claims Perplexity "disguised Comet as a Google Chrome browser" and refused to identify when the AI agent was operating on behalf of users, creating security risks and degrading the shopping experience (Reuters: 11/4/2025).

Perplexity denies wrongdoing, framing the case as anti-competitive and asserting that Amazon is using "legal threats and intimidation to stifle innovation and make life worse for consumers" (Reuters: 11/4/2025). The startup maintains that user credentials are stored locally and never on its servers, and that "simplified shopping means more transactions and happier customers."

This lawsuit highlights a critical governance gap that the industry hasn't fully addressed: agentic tools acting autonomously across third-party platforms. Platform terms of service, bot detection systems, and content protections were designed for human users and traditional bots, not AI agents making purchasing decisions on behalf of consumers (PC Mag: 11/4/2025).

The case could set important precedents affecting all agentic commerce, customer service, and productivity tools. We're likely to see the evolution of "agent access" protocols similar to robots.txt but designed for autonomous actions rather than just web crawling (The Guardian: 11/5/2025).

For companies building or deploying agentic AI, the implications are significant. You should implement explicit consent flows, session scoping, platform-specific guardrails, and auditable logs to mitigate legal and security risks. Transparency about when an agent is acting on behalf of a user, how it authenticates, and what actions it can take will likely become regulatory requirements in the near future.

As 2025 is increasingly being called the "year of agents," the absence of clear governance frameworks creates both innovation opportunities and legal landmines.

Major Tech Partnerships Reshape Consumer AI

Apple Plans to Power Siri with Google's Gemini Model

Apple is finalizing a deal to pay approximately $1 billion annually to license a custom 1.2 trillion-parameter version of Google's Gemini model to power a revamped Siri (Bloomberg, The Verge, 9to5Mac: 11/4–5/2025). The Gemini model will handle Siri's summarization and planning functions, which synthesize information and execute complex, multi-step tasks, while some Siri features will continue using Apple's in-house models (The Verge: 11/5/2025).

Critically, the custom Gemini model will run on Apple's Private Cloud Compute servers, ensuring user data never leaves Apple's infrastructure or is shared with Google (9to5Mac: 11/4/2025). This is distinct from Apple's existing ChatGPT integration, which requires user permission for each query. Gemini will be embedded within Siri's core architecture, providing seamless access to advanced reasoning without explicit opt-in prompts (The Verge: 11/5/2025).

Apple's decision follows delays in its AI roadmap and reflects CEO Tim Cook's concerns about the pace of internal AI development. The 1.2 trillion-parameter Gemini model is roughly eight times larger than Apple's current 150 billion-parameter system (The Street: 11/6/2025).

For Apple, this is a rare acknowledgment that even companies with massive resources and world-class engineering teams benefit from strategic partnerships to rapidly close capability gaps. The company is essentially trading pride for progress after years of Siri falling behind competitors like Alexa, Google Assistant, and ChatGPT (The Street: 11/6/2025).

Users should expect improved planning, summarization, context retention, and multimodal functions in Siri when the revamped assistant launches in spring 2026. The partnership may also accelerate competitive responses from Amazon and Microsoft in voice and assistant experiences (Bloomberg: 11/5/2025).

The deal underscores a broader industry trend: collaboration among tech giants is increasing even as they compete, particularly when speed-to-market and user experience are at stake.

Meta's AI Glasses Get Major Upgrades and India Launch

Meta rolled out firmware v19.2 for Ray-Ban Meta Gen 1 and Oakley Meta HSTN smart glasses, bringing significant video recording stability improvements (Android Central: 11/3/2025). Users can now select from auto, low, medium, and high stability options, with Ray-Ban Meta Gen 1 stability now on par with newer Gen 2 glasses, effectively delivering a major free upgrade to older hardware (Android Central: 11/3/2025).

Meta is also launching Ray-Ban Meta Gen 1 smart glasses in India on November 21, available on Amazon, Flipkart, and Reliancedigital.in (Times of India, Gadgets 360: 11/5–6/2025). The Indian launch includes localized features: Hindi language support for Meta AI, celebrity voice integration with Bollywood star Deepika Padukone, a festive "Restyle" feature for photos, and upcoming UPI Lite payment support allowing users to scan and pay for transactions under ₹1,000 by saying "Hey Meta, scan and pay" (Times of India: 11/6/2025).

These updates demonstrate Meta's commitment to wearable AI as a core platform alongside headsets and social apps. The combination of hardware upgrades, market expansion, and localized features signals steady progress toward mainstream adoption of "always-on" capture and assist devices (Times of India: 11/6/2025).

For enterprise applications, field work, support, translation, and training workflows continue to be early winners. However, organizations deploying these devices must address data governance challenges for always-on capture contexts, particularly around privacy, consent, and data retention.

Meta's historically strong track record of supporting older hardware with meaningful software improvements extends product lifecycles and builds user loyalty, a strategy that contrasts with some competitors' approach of reserving best features for newest devices.

Anthropic Expands European Presence with Paris and Munich Offices

Anthropic announced plans to open new offices in Paris and Munich, expanding its European footprint alongside existing hubs in London, Dublin, and Zurich (Reuters, Euronews: 11/6–7/2025). The move follows recent office openings in Tokyo, Seoul, and Bengaluru as part of a global expansion strategy that has tripled the company's international headcount over the past year.

EMEA has become Anthropic's fastest-growing region, with run-rate revenue growing more than 9x year-over-year and the number of large business accounts (representing over $100,000 in run-rate revenue each) growing more than 10x (Anthropic: 11/6/2025). Germany and France rank among the top 20 countries globally in Claude usage per capita, with European clients including L'Oréal, BMW, SAP, Sanofi, and N26 (Reuters: 11/7/2025).

Anthropic Managing Director of International Chris Ciauri stated, "Europe is home to some of the world's most important and forward-thinking companies. The business leaders I speak to are clear-eyed on both the immense opportunity that AI development represents and the critical importance of safety, reliability, and public trust" (Anthropic: 11/6/2025).

The deeper EU presence enables closer collaboration with customers, regulators, and talent as general-purpose AI rules unfold under the AI Act. Expect more region-specific partnerships, language support, and sector focus in automotive, finance, pharma, and manufacturing (Wall Street Journal: 11/7/2025).

Anthropic's emphasis on safety and transparency resonates with European enterprise buyers prioritizing ethical AI, differentiating the company from competitors like OpenAI in a market where regulatory compliance and trust are paramount.

Workforce Transformation Accelerates

Microsoft Signals AI-First Hiring After Record Layoffs

Microsoft CEO Satya Nadella announced the company will resume hiring after cutting approximately 15,000 jobs in 2025, but with a fundamentally different approach: "We will grow our headcount, but the way I look at it is that headcount will grow with a lot more leverage than what we had pre-AI" (CNBC, Economic Times: 11/1–2/2025).

Nadella emphasized that new hires will be expected to maximize productivity using AI-powered tools like Microsoft 365 Copilot and GitHub Copilot, which leverage models from OpenAI and Anthropic. "Right now, any planning, any execution, starts with AI. You research with AI, you think with AI, you share with your colleagues," Nadella said, adding that the "unlearning and learning process will take the next year or so. Then the headcount growth will come, with maximum leverage" (CNBC: 11/1/2025).

He cited an example of a network operations executive who automated DevOps maintenance using AI-powered agents rather than hiring additional staff, scaling operations without expanding headcount (Economic Times: 11/2/2025).

This "fewer people, more impact" model is becoming standard across Big Tech. As I explored in AI and Jobs: What's Really Changing and What Comes Next, the shift isn't happening at the job level; it's happening at the task level. AI is snapping onto the small but essential pieces that make up roles: summarizing, triaging, drafting, quality-checking, routine analysis. Entry-level workers who once proved themselves through these tasks are finding the training wheels of professional life removed.

Demand is shifting toward AI-augmented roles including automation design, prompt and agent operations, and data stewardship. HR and operations leaders need to update role design, training curricula, and productivity baselines to reflect AI-amplified workflows (Business Chief: 11/3/2025).

The message to the workforce is stark: employees who don't master AI tools risk obsolescence. Companies are no longer just looking for domain expertise; they're looking for domain expertise amplified by AI fluency.

October Layoffs Hit 20-Year High, AI Cited as Major Factor

U.S.-based employers announced 153,074 job cuts in October, the highest for any October since 2003 and up 175% year-over-year (Reuters, Challenger Gray & Christmas: 11/5–6/2025). Year-to-date cuts through October reached 1,099,500, a 65% increase from 2024 and the highest level since 2020 (Challenger Gray & Christmas: 11/5/2025).

The technology sector led with 33,281 October cuts, while warehousing accounted for 47,878. "Cost-Cutting" was the top reason cited, responsible for 50,437 layoffs, while "Artificial Intelligence" was the second-most cited factor, driving 31,039 cuts in October alone and 48,414 year-to-date as companies restructure and automate (Challenger Gray & Christmas: 11/5/2025).

Andy Challenger, chief revenue officer of Challenger, Gray & Christmas, stated: "October's pace of job cutting was much higher than average for the month. Some industries are correcting after the hiring boom of the pandemic, but this comes as AI adoption, softening consumer and corporate spending, and rising costs drive belt-tightening and hiring freezes. Those laid off now are finding it harder to quickly secure new roles, which could further loosen the labor market" (Reuters: 11/6/2025).

The paradox at the heart of this week's developments is striking. As detailed in my article How $125 Billion in AI Spending Becomes 75,000 Layoffs, hyperscalers are committing unprecedented capital to AI infrastructure while simultaneously cutting tens of thousands of jobs. Microsoft alone is spending $80 billion on infrastructure in fiscal 2025 while laying off 15,000 workers. The four largest tech companies are projected to spend over $360 billion on AI infrastructure in 2025 while collectively announcing 75,000+ layoffs.

This isn't hypocrisy. It's reallocation. Capital is flowing toward the infrastructure that enables AI (data centers, chips, power), while labor budgets are being restructured around AI-amplified productivity. The workers being let go aren't being replaced by AI directly; their tasks are being redistributed to fewer workers using AI tools.

A new bipartisan Senate bill, the "AI-Related Job Impacts Clarity Act," would require companies and federal agencies to submit quarterly reports of "AI-related job effects," aiming to provide transparency as AI reshapes employment (CNBC, The Register: 11/5/2025).

However, some experts warn that companies may be using AI as a rationale for layoffs while masking traditional cost-cutting and business missteps. Evidence suggests AI implementation is "exceedingly intricate and time-consuming," not the simple, rapid cost-saver portrayed by some executives (Fast Company: 11/7/2025).

The workforce transformation reflects a fundamental recalibration of how companies balance human talent with AI capabilities, with profound implications for career development, corporate culture, and labor policy.

Research Breakthroughs Push Scientific Frontiers

Google Achieves Verifiable Quantum Advantage with 13,000x Speedup

Google Quantum AI demonstrated the first verifiable quantum advantage on hardware with its "Quantum Echoes" algorithm, which ran 13,000 times faster than the world's fastest classical supercomputer, Frontier (The Quantum Insider, Google Blog: 10/21/2025). The experiment used Google's 65-qubit Willow chip to measure a quantum interference phenomenon called second-order out-of-time-order correlator (OTOC(2)), completing in just over two hours what would have taken the Frontier supercomputer approximately 3.2 years (The Quantum Insider: 10/21/2025).

The breakthrough extends beyond computational speed. Google demonstrated how the Quantum Echoes algorithm could extend nuclear magnetic resonance (NMR) spectroscopy capabilities, creating a "longer molecular ruler" that allows researchers to see interactions between atomic spins separated further apart than traditional NMR can detect (Google Blog: 10/21/2025).

Hartmut Neven, Vice President of Engineering at Google, stated: "Within five years we will see real-world applications that are only possible on quantum computers, such as quantum-enhanced sensing" (The Quantum Insider: 10/21/2025).

For enterprises, quantum computing is moving from theoretical promise to practical advantage in specialized, physically meaningful tasks. Companies should monitor developments in quantum-enhanced sensing, materials discovery, and drug design, as these capabilities are approaching commercial readiness.

AI Translates Brain Activity Into Text with "Mind-Captioning"

Researchers developed a "mind-captioning" AI system that translates brain activity into descriptive sentences using non-invasive fMRI imaging, marking a significant advance in brain-computer interfaces (Nature, Scientific American: 11/4–6/2025). Published in Science Advances, the technique uses deep-language AI models to analyze brain scans and predict "meaning signatures" while participants watch videos or imagine scenes, then generates sentence descriptions of what people are seeing or picturing in their mind (Nature: 11/4/2025).

Alex Huth, a computational neuroscientist at UC Berkeley, noted the model predicts what a person is looking at "with a lot of detail. This is hard to do. It's surprising you can get that much detail" (Nature: 11/4/2025). Beyond its scientific appeal, the technology could help people with language difficulties caused by strokes or other conditions to better communicate (Scientific American: 11/6/2025).

This opens new research directions in understanding perception, consciousness, and communication, with near-term applications in assistive technology and brain-computer interfaces.

AI Discovers New Cancer Therapy Pathway, Validated in Living Cells

Google DeepMind and Google Research collaborated with Yale to create Cell2Sentence-Scale (C2S-Scale), an AI model based on Google's Gemma family that discovered a novel, experimentally validated cancer therapy pathway (Google Blog: 10/14/2025). The model generated a hypothesis that combining silmitasertib (a drug targeting protein CK2) with low-dose interferon could increase antigen presentation by roughly 50%, making "cold" tumors more visible to the immune system and potentially more responsive to immunotherapy (Google Blog: 10/14/2025).

Remarkably, C2S-Scale's in silico prediction was confirmed multiple times in vitro in living cells, demonstrating that scaling laws can create predictive models of cellular behavior powerful enough to run high-throughput virtual screens and discover context-conditioned biology. Teams at Yale are now exploring the mechanism and testing additional AI-generated predictions in other immune contexts (Google Blog: 10/14/2025).

AI-driven hypothesis generation and virtual screening are compressing drug discovery timelines from years to months, with experimentally validated results demonstrating the technology's readiness for real-world application.

Nobel Laureate's Lab Achieves AI-Designed Antibody Breakthrough

Nobel laureate David Baker's lab achieved another breakthrough with AI-designed antibodies that successfully hit their intended targets, marking a milestone in protein design (GeekWire, Drug Target Review: 11/6–7/2025). The advance demonstrates AI's growing capability to design functional biological molecules entirely from scratch, with implications for therapeutics, diagnostics, and materials science.

These breakthroughs underscore AI's expanding role not just in software and services but as a fundamental tool for scientific discovery and innovation across domains.

Enterprise Adoption Reaches Critical Mass

82% of Leaders Use AI Weekly; ROI Tracking Becomes Standard

The 2025 Wharton-GBK Collective AI Adoption Report reveals that 82% of enterprise leaders now use generative AI at least weekly, up from 72% in 2024, with 46% using it daily, a 17-point year-over-year increase (LinkedIn, Knowledge@Wharton: 11/1/2025). The study shows enterprises have transitioned from exploration (2023) and experimentation (2024) to "accountable acceleration" (2025), with 72% now formally tracking Gen AI ROI focusing on productivity gains and incremental profit (Knowledge@Wharton: 10/28/2025).

Three out of four leaders report positive returns on Gen AI investments, and 88% plan to increase AI budgets in the next 12 months. One-third of AI budgets are now allocated to internal R&D, signaling a shift toward custom-built solutions rather than generic tools (Knowledge@Wharton: 10/28/2025).

Top use cases include data analysis (73%), document summarization (70%), and marketing content creation (66%). CAIO (Chief AI Officer) roles now exist in 60% of enterprises, up sharply year-over-year, with 67% of executive teams directly engaged in AI strategy (Knowledge@Wharton: 10/28/2025).

Fortune 500 Adoption Triples Year-Over-Year

As of October 2025, 67 Fortune 500 companies have deployed an enterprise LLM product to their employees (13.4% of Fortune 500), more than tripling from just 22 companies in October 2024 (Bloomberry: 11/1/2025). Wavestone's Global AI Survey 2025 found that 90% of companies now include AI in their business strategy, with an average of 13% of IT budgets allocated to AI (Wavestone: 10/28/2025).

The era of experimental pilots is ending. Organizations are now expected to demonstrate measurable ROI, productivity gains, and bottom-line impact from AI investments. As adoption accelerates, the talent gap is becoming the primary bottleneck, requiring aggressive investment in training, upskilling, and recruiting AI-fluent talent (Wavestone: 10/28/2025).

With 90% strategic integration and growing regulatory scrutiny, robust governance frameworks for data, ethics, security, and transparency are no longer optional.

AI Startups Capture Record 53% of Global VC Funding

AI startups captured a projected record $192.7 billion in global venture capital funding in 2025, surpassing 53% of all VC funding worldwide (Economic Times: 10/21/2025). This marks the first time AI has captured more than half of total VC investment.

The surge is dominated by a handful of major firms including Anthropic and xAI, which have attracted multi-billion-dollar rounds, creating a capital concentration dynamic where AI startups capture most funding while other sectors struggle to access capital (Economic Times: 10/21/2025).

This winner-take-most dynamic means startups outside the AI sector face significant capital access challenges, while AI incumbents enjoy unprecedented funding availability.

What This Week Tells Us

Compute access is becoming a strategic differentiator, not just an operational input. OpenAI's $38 billion AWS deal, VAST-CoreWeave's $1.17 billion partnership, and Anthropic's commitment to one million Google TPUs demonstrate that AI leaders are locking in long-term, integrated infrastructure stacks (GPUs, data platforms, and energy) to power faster features, stable performance, and market leadership. As I outlined in The $7.8 Trillion Bet, this represents the largest coordinated infrastructure buildout since rural electrification, committed before clear monetization pathways exist for most use cases.

Regulation is caught between innovation and control. The EU's proposed AI Act delays and grace periods reflect mounting pressure from U.S. authorities and Big Tech, offering organizations breathing room while maintaining long-term compliance expectations. However, fragmented U.S. state-level rules and geopolitical chip restrictions create persistent complexity.

Geopolitics is tightening chip supply and bifurcating markets. Nvidia's firm stance on not selling Blackwell chips to China, combined with enduring U.S. export controls, ensures ongoing GPU scarcity, regional supply fragmentation, and accelerated domestic chip development in China. Enterprises need multi-vendor accelerator strategies and hybrid inference approaches.

Agents are forcing new governance frameworks. The Amazon-Perplexity lawsuit signals emerging standards for agent identity, permissions, and platform compliance. Expect policies governing autonomous actions comparable to robots.txt, but for purchasing, booking, and other complex behaviors rather than just web crawling.

Consumer AI is marching toward ubiquity. Apple's Gemini-powered Siri, Meta's Ray-Ban glasses updates, and expanding global launches show steady progress in making multimodal assistance mainstream. Watch for new privacy expectations, provenance requirements, and governance frameworks for ambient capture.

Workforce transformation is accelerating, and it's painful. Microsoft's "AI-first" hiring philosophy and October's 20-year-high layoffs underscore the rapid recalibration of talent strategy. As I detailed in AI and Jobs: What's Really Changing, the economy-wide "jobs shock" hasn't arrived yet, but the ladders people climb to build careers are already losing rungs. Entry-level roles that once taught judgment, discipline, and context are being quietly automated away. Organizations must urgently upskill existing employees, redesign roles for AI leverage, and navigate the tension between automation gains and workforce disruption.

Enterprise adoption has crossed the chasm. With 82% weekly usage, 72% formal ROI tracking, and Fortune 500 adoption tripling year-over-year, AI has moved from experimentation to operational integration. The imperative now is proving performance at scale, not pilot success.

The next few months will be critical. Watch for EU AI Act decisions in mid-November, continued chip geopolitics, evolving agent governance frameworks, and workforce policy responses to AI-driven displacement. The infrastructure deals signed this week will shape competitive dynamics for years to come.

References

Infrastructure & Compute

Reuters (11/3/2025): OpenAI turns to Amazon in $38 billion cloud services deal

CNBC (11/3/2025): Amazon closes at record after $38 billion OpenAI deal with AWS

Wired (11/3/2025): OpenAI Signs $38 Billion Deal With Amazon

Reuters (11/6/2025): Nvidia-backed Vast Data inks $1.17 billion AI deal with CoreWeave

Blocks and Files (11/5/2025): VAST Data lands $1.17 billion CoreWeave deal

VentureBeat (11/6/2025): Google debuts AI chips with 4X performance boost

Economic Times (11/5/2025): Google to offer Ironwood TPU for public use

Policy & Regulation

Financial Times (11/7/2025): EU set to water down landmark AI act after Big Tech pressure

Fortune (11/7/2025): EU considers weakening landmark AI Act

Wall Street Journal (11/7/2025): EU Floats Tweaks, New Grace Periods in AI Act

Reuters (11/6/2025): Nvidia CEO says no active discussions on selling Blackwell chip to China

Bloomberg (11/7/2025): Nvidia CEO Says No Plans to Ship Blackwell AI Chips to China

CNBC (11/6/2025): China's key weapons in its AI battle with the U.S.

Agentic AI & Governance

Reuters (11/4/2025): Amazon sues Perplexity over 'agentic' shopping tool

PC Mag (11/4/2025): Amazon Sues Perplexity As AI Browser War Escalates

The Guardian (11/5/2025): Amazon sues AI startup over browser's automated shopping feature

Tech Partnerships

Bloomberg (11/5/2025): Apple Nears $1 Billion-a Year Deal to Use Google AI for Siri

The Verge (11/5/2025): Apple is planning to use a custom version of Google Gemini

9to5Mac (11/4/2025): Apple nears $1 billion Google deal for custom Gemini model

Android Central (11/3/2025): Ray-Ban Metas video recording upgrade

Times of India (11/6/2025): Meta's AI-powered Ray-Ban glasses India launch

Reuters (11/7/2025): AI startup Anthropic expands in Europe

Anthropic Blog (11/6/2025): New offices in Paris and Munich

Workforce Transformation

CNBC (11/1/2025): Microsoft will grow headcount with more leverage

Economic Times (11/2/2025): Nadella says Microsoft will hire again with AI-first approach

Reuters (11/6/2025): US layoffs for October surge to two-decade high

Challenger Gray & Christmas (11/5/2025): October Challenger Report

CNBC (11/5/2025): New bipartisan bill would require companies to report AI job losses

Fast Company (11/7/2025): AI isn't replacing jobs. AI spending is

Research Breakthroughs

The Quantum Insider (10/21/2025): Google Quantum AI Shows 13,000x Speedup

Google Blog (10/21/2025): The Quantum Echoes algorithm breakthrough

Nature (11/4/2025): Mind-captioning AI decodes brain activity

Scientific American (11/6/2025): AI Decodes Visual Brain Activity

Google Blog (10/14/2025): Google's Gemma AI model cancer discovery

GeekWire (11/6/2025): Nobel winner's lab AI-designed antibodies

Enterprise Adoption

Knowledge@Wharton (10/28/2025): 2025 AI Adoption Report

Bloomberry (11/1/2025): The state of AI adoption in large orgs

Wavestone (10/28/2025): Global AI survey 2025

Economic Times (10/21/2025): AI startups captured over 50% of venture funding

Related Analysis by Jay Cadmus

How $125 Billion in AI Spending Becomes 75,000 Layoffs

AI and Jobs: What's Really Changing and What Comes Next

The $7.8 Trillion Bet: What Unprecedented AI Infrastructure Investment Reveals