Alphabet’s AI chips are a potential $900 billion ‘secret sauce’

Alphabet’s TPU chips are driving growing investor optimism that the company can create a significant new revenue stream: TPUs helped spark a roughly 30% Q4 stock rally and deals — including a multi‑billion dollar supply pact with Anthropic and reported talks with Meta — highlight third‑party demand. Morgan Stanley and other analysts estimate TPU unit demand rising to ~5 million in 2027 and 7 million in 2028, and note that every 500,000 TPUs sold externally could add about $13 billion to 2027 revenue (roughly a 3% lift against a projected $447 billion revenue) and ~$0.40 in EPS. Risks include uncertainty around external sales execution and a stretched valuation (~27x forward earnings), but the combination of Gemini optimization, cloud integration and potential ASIC cost advantages versus Nvidia underpin positive revenue upside scenarios.

Analysis

Market structure: TPUs shift marginal share away from general-purpose GPU incumbents (NVDA) toward vertically integrated ASIC suppliers (GOOGL/GOOG), benefiting Google Cloud, Anthropic, and potentially Meta if they buy access; Morgan Stanley’s math (every 500k external TPUs ≈ $13bn revenue, ~3% of 2027 sales) frames a realistic near‑term upside. Supply/demand: GPU scarcity + high Nvidia pricing creates a 12–36 month window for TPU adoption, but build capacity (foundry/OSAT) and software/tooling will cap pace. Cross-asset: stronger AI clarity pushes risk-on — equity risk premia compress, IG spreads tighten; NVDA options vol likely to fall if market pricing shifts toward ASICs.

Risk assessment: Tail risks include failed third‑party commercialization, exclusive deals that limit addressable market, foundry bottlenecks, and antitrust/regulatory action if Google ties TPU access to Cloud — low probability but high impact. Timeline: immediate (days) = sentiment swings on partnership headlines; short (3–12 months) = proof points via 3rd-party contracts; long (2025–2028) = revenue/margin impact if >1–2m external TPUs shipped. Hidden dependencies: software migration costs, Gemini ecosystem lock‑in, and exclusivity clauses could materially change adoption curves.

Trade implications: Direct: establish a 2–3% long position in GOOGL (GOOGL) now, scaling to 4–6% if Google announces ≥2 non‑exclusive TPU customer contracts within 6 months or Morgan Stanley TPU shipment revisions persist upward. Pair: long GOOGL / short NVDA (NVDA) notional 0.6x over 12–24 months to express ASIC share gain while limiting single‑name beta. Options: buy a Jan 2026 GOOGL call spread (buy 1.2× ATM, sell 1.5× ATM) sized to 1–2% portfolio risk and buy NVDA 9–12 month 10–15% OTM puts as hedge.

Contrarian angles: Consensus assumes rapid third‑party TPU adoption; that ignores multi‑year porting costs and potential foundry limits — a Graviton‑style ramp took AWS several years to dent incumbents. Reaction may be overdone: market-implied 20% AI share is aggressive absent broad software portability and non‑exclusive supply. Watch for unintended consequences: antitrust scrutiny or customer anti‑lockin clauses that could materially cap external TPU revenue; treat any >$5bn third‑party TPU revenue guide within 12 months as definitive bullish signal.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in Alphabet (GOOGL) now; increase to 4–6% only if Google announces at least two non‑exclusive third‑party TPU contracts within the next 6 months or if broker TPU shipment estimates rise by >50% year-over-year.
• Initiate a relative‑value pair: long GOOGL / short NVDA (NVDA) at 0.6x notional exposure, 12–24 month horizon, to express TPU-driven share gains while hedging GPU market risk; cap pair to 2% net portfolio delta.
• Purchase a Jan 2026 GOOGL call spread (buy ~1.2× ATM, sell ~1.5× ATM) sized to risk 1–2% of portfolio to capture multi‑year upside without overpaying for volatility; simultaneously buy NVDA 9–12 month 10–15% OTM puts (size 25–40% of GOOGL notional) as downside insurance against GPU repricing shocks.
• Trim GOOGL exposure to 1–2% if forward P/E >30x or if quarterly guidance omits any third‑party TPU revenue; conversely, add if management discloses >$5bn of third‑party TPU revenue guidance or analysts raise 2027 revenue estimates by >5% linked explicitly to TPUs.