AI Market Intelligence Monthly Digest — 2026-05-31

Executive Summary

Anthropic overtook OpenAI as the world's most valuable AI startup, closing a $30 billion round at a $900+ billion valuation while poaching founding OpenAI researcher Andrej Karpathy — a twin blow to its primary rival that reshapes the competitive landscape. Bloomberg
Big Tech AI capital expenditure commitments reached $725 billion for 2026, with Microsoft alone guiding to $190 billion and Meta planning up to $135 billion, while a landmark $25 billion Google-Blackstone joint venture signals private capital flooding into AI infrastructure at unprecedented scale. Tom's Hardware
Custom AI chips are outpacing NVIDIA GPU growth for the first time, with ASIC shipments projected to grow 44.6% versus 16.1% for merchant GPUs — a structural shift compounded by HBM memory shortages that have locked up supply into 2027 and are driving consumer electronics prices up 15–20%. TechTimes
The EU reached a political agreement to delay key AI Act compliance deadlines to late 2027 and 2028 while adding new prohibitions on deepfake-nude and CSAM generation tools; in the US, President Trump abruptly pulled a planned AI executive order, leaving federal AI governance in limbo. European Council
NVIDIA posted record Q1 FY2027 revenue of $81.6 billion (85% YoY growth), but investor concerns over mounting competition from custom ASICs and a forward-looking guidance miss sent mixed signals through the market — even as its market cap surpassed $5.2 trillion. CNBC

AI Industry News

Anthropic's Twin Coup: Record Valuation and Karpathy Hire

May 2026 was dominated by Anthropic's rapid consolidation of competitive advantage. On May 22, Bloomberg reported that Anthropic was closing a funding round exceeding $30 billion at a valuation above $900 billion, co-led by Sequoia Capital, Dragoneer Investment Group, Altimeter Capital, and Greenoaks Capital Partners, each investing roughly $2 billion. Bloomberg. This is Anthropic's second $30 billion raise in 2026 alone — the first, in February, valued the company at $380 billion — meaning Anthropic's valuation has more than doubled in approximately three months.

Three days earlier, Andrej Karpathy — a founding member of OpenAI and former leader of Tesla's Autopilot team — announced he was joining Anthropic's pre-training team, where he will help launch a new initiative using Claude itself to accelerate pretraining research. Axios. The move is widely viewed as the most consequential AI talent shift of the year and a significant blow to OpenAI's reputational standing among elite researchers.

Google I/O 2026: Gemini 3.5 Flash and Pricing Moves

At Google I/O on May 19, Google launched Gemini 3.5 Flash, described as the first in its latest series of models combining frontier intelligence with agentic action. The model outperforms Gemini 3.1 Pro on coding and agentic benchmarks including Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo), and MCP Atlas (83.6%). Gemini 3.5 Flash is generally available via Google Antigravity, the Gemini API in Google AI Studio, and Android Studio. Google also cut its Ultra subscription from $250 to $200/month, a clear pricing play to maintain competitive positioning against OpenAI and Anthropic. Google Blog

OpenAI Pivots Toward Revenue Diversification

OpenAI made two strategically significant moves in May. First, the company launched a self-serve Ads Manager platform enabling advertisers to create, manage, and optimize campaigns inside ChatGPT. The platform reportedly targets $2.5 billion in ad revenue this year and $100 billion annually by 2030, and supports both CPM and CPC buying models with integrations across Dentsu, Omnicom, Publicis, WPP, Adobe, Criteo, and StackAdapt. MarketingProfs

Second, OpenAI launched the OpenAI Deployment Company, a $4 billion enterprise services unit that embeds Forward Deployed Engineers directly inside customer organizations. The division acquired AI consulting firm Tomoro (~150 deployment specialists) to seed the operation, putting OpenAI in direct competition with traditional consulting firms like Accenture and Deloitte. CXtoday

IBM Think 2026 and Enterprise AI

IBM unveiled its most comprehensive enterprise AI expansion at Think 2026 on May 5, including next-generation watsonx Orchestrate for multi-agent orchestration, IBM Confluent for real-time data integration, IBM Concert for intelligent operations, and IBM Sovereign Core for operational independence. IBM Newsroom

Hardware, Datacenter & Energy

The Custom Chip Inflection Point

Perhaps the most structurally significant hardware development this month: custom AI chips (ASICs) are outpacing NVIDIA GPU shipment growth for the first time. TrendForce projects 44.6% ASIC growth versus 16.1% for merchant GPUs in 2026, with ASIC-based AI server shipments reaching 27.8% of the total AI server market. TechTimes. This shift reflects hyperscalers' strategic investment in purpose-built silicon (Google TPUs, Amazon Trainium, Meta's MTIA) to reduce dependence on NVIDIA and optimize cost-per-inference for their specific workloads.

NVIDIA: Record Revenue, Mounting Concerns

NVIDIA's Q1 FY2027 results on May 20 delivered headline-beating numbers — $81.6 billion in revenue (85% YoY), data center revenue of $39.1 billion (+69% YoY), and record net income of $58.3 billion. CNBC. However, Bloomberg reported that the forward guidance disappointed investors as competition from custom ASICs mounts. NVIDIA signaled that governments and non-hyperscale enterprises would need to become bigger revenue sources as its customer base diversifies. Bloomberg

GPU and Memory Supply Crisis Deepens

GPU shortages persisted throughout May with Blackwell pricing up 15–23% and lead times extending to 3–7 months. MSI confirmed a 20% GPU supply gap from NVIDIA that could shrink the PC market by as much as 20%. Fusion Worldwide; TweakTown

The root constraint remains HBM (High-Bandwidth Memory), which has reached crisis-level scarcity. Each AI GPU requires eight HBM modules consuming 96 DRAM dies, and AI firms have locked up supply well into 2027. HBM now consumes 23% of total DRAM wafer output, up from 19% last year, crowding out commodity memory and driving 15–20% PC price hikes from Lenovo, Dell, HP, Acer, and ASUS. Fortune; Bloomberg

Datacenter Megaprojects Accelerate

The Google-Blackstone $25 billion AI cloud infrastructure joint venture, announced May 19, will target 500 megawatts of computing capacity by 2027 powered by Google's TPUs. Benjamin Treynor Sloss will serve as CEO. The deal represents a new model for datacenter financing — blending hyperscaler AI expertise with institutional capital at scale to challenge neocloud providers. CNBC

Microsoft confirmed it raised calendar 2026 capex to $190 billion, with $25 billion attributed to cost increases in DRAM, flash, wafers, and substrates. The company added 1 GW of capacity in Q1 and remains on track to double capacity in two years. NextPlatform

Meta's AI spending plans of up to $135 billion in 2026 include its "Hyperion" datacenter campus in Louisiana, which could draw up to 5 GW of power — roughly equivalent to 4.2 million homes. SEC Filing. The sheer energy scale of these commitments is raising increasingly urgent questions about grid capacity and renewable energy timelines.

Financial & Deal Flow

Earnings Season Highlights

NVIDIA dominated May earnings with Q1 FY2027 revenue of $81.6 billion and EPS of $1.87, both beating consensus ($78.8B / $1.76). Market cap crossed $5.2 trillion. CNBC

AMD reported Q1 2026 revenue of $10.3 billion (+38% YoY), with strong growth driven primarily by data center and client segments, underscoring the breadth of AI chip demand beyond NVIDIA. AMD SEC Filing

Baidu noted that its core AI-powered business exceeded 50% of total general business revenue for the first time in Q1 2026, a milestone for a major Chinese tech company. Baidu SEC Filing

SoundHound AI posted Q1 revenue of $44.2 million (+52% YoY), guiding to $225–$260 million for full-year 2026. Motley Fool

Palladyne AI, a defense-focused AI company, reported Q1 revenue up 107% YoY to $3.5 million with $17 million backlog and full-year guidance of $24–$27 million (357–415% growth). Palladyne SEC Filing

Major Funding Rounds

Company	Amount	Valuation	Lead Investors	Date
Anthropic	$30B+	$900B+	Sequoia, Dragoneer, Altimeter, Greenoaks	May 22
Cognition AI	$1B	$26B	Lux Capital, General Catalyst, 8VC	May 27
Google-Blackstone JV	$25B (total)	N/A	Blackstone ($5B equity)	May 19

Cognition AI, maker of the Devin AI software engineer, raised over $1 billion at a $26 billion valuation, more than doubling from its September round. Lux Capital, General Catalyst, and 8VC co-led with participation from Ribbit Capital, Atreides Management, and Peter Thiel's Founders Fund. Bloomberg

Capital Expenditure Arms Race

The hyperscaler capex aggregate for 2026 has reached staggering proportions: Google, Amazon, Microsoft, and Meta collectively plan $725 billion, up 77% from $410 billion in 2025. AI-related infrastructure alone is expected to exceed $600 billion. Tom's Hardware; Fortune. Microsoft stated it expects to remain capacity-constrained through 2026, implying demand continues to outstrip even this record spending.

Policy & Regulation

EU AI Act: Deadlines Extended, Scope Expanded

The most consequential regulatory development in May was the EU's "AI Omnibus" provisional agreement on May 7. The Council and Parliament agreed to delay key compliance deadlines: standalone high-risk AI systems (Annex III — employment, education, biometrics, critical infrastructure, migration) now have until December 2, 2027, while high-risk AI embedded in products (Annex I) is pushed to August 2, 2028. Crucially, a new prohibition was added to Article 5 targeting AI systems used to generate non-consensual sexual/intimate images ("nudification" tools) and CSAM. European Council

On May 19, the European Commission published draft guidelines on high-risk AI classification, clarifying that intended purpose is determined holistically — a provider cannot escape high-risk classification merely by excluding high-risk uses in its terms of service if product positioning tells a different story. Public consultation is open until June 23, 2026. Hunton Privacy Blog

US Federal AI Governance in Disarray

On May 21, President Trump abruptly canceled the signing of a landmark executive order on AI oversight, reportedly because it could "interfere with American competitiveness." The order would have formalized cybersecurity standards and a voluntary testing regime for frontier AI models, directed by CISA, the National Cyber Director, and OSTP. Its cancellation signals deep internal White House divisions on the appropriate level of AI governance and leaves US federal AI policy in significant limbo. NBC News

Meanwhile, the FTC began enforcing the TAKE IT DOWN Act, requiring platforms to remove non-consensual intimate images (including AI-generated deepfakes) within 48 hours of receiving a request. FTC. Separately, the FTC settled with Cox Media Group for nearly $1 million over deceptive claims about an "active listening" AI-powered marketing service, continuing the agency's aggressive "Operation AI Comply" enforcement initiative. FTC

UK Criminalizes Harmful AI Content Generation

The UK Crime and Policing Act 2026, which received Royal Assent on April 29, creates new criminal offences for AI tools optimized to generate CSAM, deepfake intimate imagery, and "purported intimate image generators," extending both corporate and individual liability. The Act also amends the Online Safety Act 2023 to empower ministers to regulate "illegal AI-generated content" and designates deepfake intimate imagery offences as priority offences, requiring platforms to proactively prevent such content. ResultSense

Market Signals & Analysis

Theme 1: The Anthropic-OpenAI Power Shift Is Real

Anthropic's $900 billion+ valuation, combined with the Karpathy hire and OpenAI's defensive pivots into advertising and consulting services, suggests a meaningful shift in the AI competitive center of gravity. OpenAI's diversification moves — ads and enterprise deployment — signal that management sees model-level differentiation eroding and is racing to build durable business moats through distribution and services. Anthropic, by contrast, is doubling down on model excellence and talent acquisition. Whether this divergence represents sustainable strategic advantage or simply different phases of the same trajectory remains to be seen, but the talent market is voting with its feet.

Theme 2: The ASIC Crossover and NVIDIA's Strategic Challenge

For the first time, custom ASIC growth is outpacing NVIDIA's merchant GPU business. This is not a sudden shift — hyperscalers have been investing in custom silicon for years — but the crossover point matters because it signals that the most capital-rich AI buyers are becoming less dependent on NVIDIA. Combined with NVIDIA's guidance disappointment and its acknowledged need to diversify beyond hyperscale customers, this theme bears close watching. NVIDIA's $81.6 billion quarter remains extraordinary by any measure, but the growth narrative is evolving.

Theme 3: Infrastructure Spending Is Outrunning Physical Constraints

The aggregate $725 billion in hyperscaler capex, Meta's 5 GW datacenter campus, and Microsoft's admission that it will remain capacity-constrained despite spending $190 billion all point to the same conclusion: the AI infrastructure buildout is pressing against hard physical limits — power, memory, and land. The HBM shortage is the most acute bottleneck today, with consequences cascading into consumer electronics pricing. The Google-Blackstone JV represents a creative financing response, but the underlying constraint is manufacturing capacity, not capital.

Theme 4: Regulatory Divergence Deepens

The transatlantic regulatory gap widened in May. The EU extended its AI Act deadlines — a pragmatic concession to industry readiness — while simultaneously adding new prohibitions and publishing detailed classification guidance. The US, by contrast, saw its most significant federal AI governance attempt collapse at the eleventh hour due to political disagreements about competitiveness. The FTC remains active on enforcement, but the absence of a coherent federal framework leaves the US without a structural counterpart to the EU's approach. The UK is carving its own path through criminal law. Companies operating globally must now navigate three substantially different regulatory philosophies.

Theme 5: AI Monetization Strategies Are Diverging Rapidly

OpenAI's ad platform launch, Anthropic's model-licensing approach, Google's aggressive subscription pricing, and the enterprise deployment services trend represent fundamentally different theories of how AI companies will capture value. The advertising play is the most ambitious and contentious — it positions ChatGPT as a search advertising competitor, but risks user trust. The enterprise services pivot echoes the cloud wars, where professional services became a critical differentiator. The coming quarters will reveal which monetization models generate sustainable unit economics.

Key Items to Watch

Anthropic round close: Confirmation of final terms and whether the $900B+ valuation holds; watch for any new product announcements timed to the close.
EU AI Act high-risk guidelines consultation: Deadline June 23 — final guidelines will determine compliance scope for thousands of companies.
US AI executive order: Whether the Trump administration revives, revises, or permanently shelves the pulled AI governance order will define federal policy direction.
NVIDIA Computex/analyst day: Watch for competitive responses to ASIC momentum and any new product announcements targeting sovereign AI or enterprise customers.
HBM supply updates: Q3 pricing projections and any capacity expansion announcements from Samsung, SK Hynix, or Micron will signal whether the shortage peaks or extends.
OpenAI advertising traction: Early metrics from the ChatGPT Ads Manager — adoption rates, advertiser mix, and user engagement impact — will test the ad-supported model thesis.
Meta Hyperion power sourcing: Any developments on the 5 GW energy sourcing for Louisiana, including renewable energy contracts or nuclear partnerships.
Cognition AI / Devin product updates: At $26B valuation, market will watch closely for enterprise adoption metrics to justify the pricing.