Alphabet’s AI Turnaround Rewrites Big Tech Investment Thesis

Alphabet’s market capitalisation briefly surpassed Nvidia at $5.2 trillion in after-hours trading on 10 May 2026, marking the culmination of a 160% stock rally driven by competitive recovery in AI and a strategic pivot from perceived laggard to full-stack infrastructure leader. The company’s transformation from anxious catch-up player to credible competitor reshapes the $10 trillion Big Tech valuation thesis: vertical integration across custom silicon, proprietary models, and cloud capacity now appears more defensible than pure model leadership as foundation model performance converges and enterprise buyers prioritise infrastructure lock-in over benchmark superiority.

The narrative inversion accelerated with Gemini 3.1 Pro leading LMArena leaderboards with 84.0% on GPQA Diamond and 92.0% on AIME 2024—scores that trade benchmarks with OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.7, according to AI Wiki. Gemini’s monthly active user base reached 750 million in Q4 2025, up from 650 million the prior quarter, per CNBC. The technical catch-up matters less than the strategic shift it enables: Alphabet no longer competes solely on model quality but on infrastructure leverage—using proprietary TPU silicon, cloud distribution, and cost advantages from Vertical Integration to create switching costs OpenAI and Anthropic cannot replicate without billions in capex they do not control.

The $190 Billion Bet on Vertical Integration

Alphabet raised 2026 capital expenditure guidance to $180-190 billion, nearly doubling the $91.4 billion spent in 2025, per Data Center Dynamics. The figure dwarfs OpenAI’s estimated compute costs and positions Alphabet as the largest AI Infrastructure spender among hyperscalers tracking a combined $650-700 billion in 2026 capex. CFO Anat Ashkenazi stated the spend would target “AI compute capacity for Google DeepMind and to meet significant cloud customer demand as well as strategic investments in other bets.” The allocation reflects conviction that owning the stack—from chip design to model training to cloud delivery—creates margin advantages competitors leasing capacity from AWS or Azure cannot match.

The TPU calculus underpins this thesis. Citizens analyst Andrew Boone projects Alphabet’s custom silicon will generate $3 billion in revenue in 2026 and $25 billion in 2027, per Fortune. While these figures remain estimates, they illustrate the long-term value proposition: Alphabet can subsidise internal compute costs below market rates while selling excess capacity to cloud customers, creating a pricing moat competitors relying on Nvidia H200 clusters cannot replicate. CEO Sundar Pichai acknowledged on the February 2026 earnings call that compute capacity constraints “keep me up at night,” expecting supply bottlenecks throughout 2026—a signal the company views infrastructure as the binding constraint on AI dominance, not model architecture.

OpenAI Stumbles, Anthropic Surges

Alphabet’s recovery coincides with deteriorating competitive dynamics for OpenAI. The Motley Fool reported the company missed internal revenue and user growth targets at the end of 2025, with CFO Sarah Friar expressing concerns about affording computing contracts. While OpenAI released GPT-5.5 on 23 April 2026 and GPT-5.5 Instant on 5 May 2026—claiming 52.5% fewer hallucinations than GPT-5.3—the improvements arrived amid growing enterprise scepticism about model-first vendors without infrastructure ownership.

Anthropric captured approximately 70% of new enterprise AI deals versus OpenAI by March 2026, with enterprise revenue surpassing OpenAI by mid-2025, according to Ramp platform analysis cited by Android Headlines. Estimates of Anthropic’s annualised revenue range from $14 billion to $30 billion as of April 2026, depending on methodology. The competitive reshuffling suggests enterprise buyers prioritise safety positioning and infrastructure partnerships over raw model performance—a dynamic favouring Alphabet’s embedded cloud strategy over OpenAI’s standalone API approach. Anthropic’s “cultural moat” around safety and its distribution partnerships with AWS and Google Cloud create switching costs that model benchmarks alone do not overcome.

Market Repricing of AI Concentration Risk

Alphabet’s valuation surge from $1.8 trillion in May 2025 to $4.8 trillion in May 2026 reflects sector-wide reassessment of who wins in a capital-intensive AI era. The company trades at 28-35x forward price-to-earnings, per TechTimes—a premium justified by Google Cloud’s growth and $462 billion backlog, but vulnerable if capex returns disappoint. Free cash flow fell 47% year-over-year to $10.1 billion in Q1 2026 despite 22% revenue growth, as capex nearly doubled while operating expenses grew 27%. The divergence raises sustainability questions: can Alphabet maintain margin expansion while absorbing $190 billion in annual infrastructure spending, or does the capex cycle reset profitability expectations across Big Tech?

The concentration thesis cuts both ways. Bulls argue Alphabet’s stack ownership insulates it from compute cost inflation and creates pricing power as smaller competitors exhaust capital. Divyaunsh Divatia of Janus Henderson Investors told Fortune that “Alphabet just has everything you want, and that’s why everyone is so comfortable owning it, because it has so many ways to win within AI.” Bears counter that elevated multiples assume sustained 60%+ cloud growth and seamless monetisation of Gemini across search, YouTube, and enterprise—outcomes vulnerable to regulatory intervention, model commoditisation, or enterprise backlash against vendor lock-in. The company’s May 2026 market cap near-parity with Nvidia suggests investors view full-stack AI infrastructure as comparably defensible to semiconductor dominance, a repricing with profound implications for tech sector capital allocation.

What to Watch

Alphabet’s Q2 2026 earnings (late July) will test whether Google Cloud’s growth rate can persist as enterprise AI spending normalises and whether Gemini monetisation across consumer surfaces (Search, YouTube, Android) justifies the $190 billion capex outlay. Three metrics matter: (1) Google Cloud operating margin trajectory—can the unit expand margins while absorbing infrastructure costs, or does hyperscale AI commoditise cloud services? (2) Gemini conversion rates in Search—does AI-enhanced search increase ad revenue per query or cannibalise traditional text ads? (3) TPU external revenue—does Alphabet successfully sell excess compute capacity to third parties, validating the vertical integration thesis, or do enterprise customers prefer Nvidia-based solutions for portability?

Regulatory developments pose asymmetric risk. The US Department of Justice’s antitrust scrutiny of Google Search intensifies as Gemini integration raises questions about self-preferencing in AI-generated results, while European Union AI Act compliance costs could reduce margin advantages from vertical integration. On the upside, any sign OpenAI cannot afford its compute commitments or Anthropic’s growth stalls would accelerate enterprise consolidation toward hyperscaler-backed models, strengthening Alphabet’s embedded distribution advantage.

The broader investment thesis hinges on whether AI infrastructure returns justify the $650-700 billion hyperscaler spending cycle. If Alphabet’s TPU silicon generates the projected $25 billion in 2027 revenue while Google Cloud sustains 50%+ growth, the full-stack model validates and competitors lacking chip design or hyperscale infrastructure face margin compression. If capex-to-revenue conversion disappoints—either through model commoditisation, regulatory constraints, or enterprise resistance to lock-in—the current valuation multiples become unsustainable and force a sector-wide repricing. Alphabet’s ability to monetise its compute advantage while maintaining innovation velocity against nimbler competitors will determine whether the company’s 160% rally represents durable competitive repositioning or peak-cycle enthusiasm for infrastructure spending that exceeds near-term demand.