Google’s AI chips 'great thing' but will not shake Nvidia: Wedbush

Wedbush commented that Google's Tensor Processing Units (TPUs) do not materially threaten Nvidia's (NVDA) position or CEO Jensen Huang, responding to reports that Meta Platforms (META) is in talks to spend billions on Alphabet's (GOOGL/GOOG) AI chips. The remark frames Google's chip initiative as a competitive development worth monitoring but not an immediate market-disrupting event for Nvidia, implying limited near-term impact on NVDA market share or investor positioning.

Analysis

Market structure: Nvidia retains near-term pricing power and share in training GPUs because of CUDA lock-in, broad OEM support and capacity constraints at TSMC — expect continued demand-driven premium for 6–12 months. Google (GOOGL/GOOG) wins optionality: TPU sales to hyperscalers/cloud customers are high-margin incremental revenue but adoption will be measured (likely <5–10% displacement of Nvidia spend in first 12–24 months). Meta’s (META) talks imply potential large-ticket procurement but mostly redistribute hyperscaler spend rather than expand overall market immediately.

Risk assessment: Tail risks include a surprise large Meta+other hyperscaler TPU procurement (>=$1bn per year) shifting multiyear demand; aggressive regulatory/antitrust action splitting software-hardware stacks; and TSMC/manufacturing outages that could flip supply dynamics. Near-term (days–weeks) risk is headline-driven IV spikes; medium-term (3–12 months) depends on procurement cadence and benchmark disclosures; long-term (2+ years) depends on software portability and standards that could erode CUDA stickiness. Hidden dependencies: model-compiler portability, total cost of ownership, and datacenter integration effort.

Trade implications: Favor tactical NVDA overweight for 3–12 months sized 2–4% portfolio with downside hedges (see decisions). Buy optional exposure to GOOGL upside from TPU monetization via 9–18 month LEAP calls sized 1–2%. Consider a relative-value pair (long GOOGL, short META) for 6–12 months to capture TPU monetization vs Meta capex uncertainty; rotate into cloud/software names if Nvidia margin guidance weakens by >200bps.

Contrarian angles: Market may overstate immediate TPU threat to Nvidia — historical analogs (Intel vs GPU cycles) show software ecosystems slow migration for 2–4 years. Conversely, the market may underprice direct revenue upside to GOOGL if TPU sales scale; a single multi-year hyperscaler contract >$2bn would re-rate GOOGL's infrastructure segment. Unintended consequence: large TPU adoption could force Nvidia to accelerate price cuts or software concessions, compressing GPU ASPs and margins faster than consensus expects.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (equities) with a 3–12 month horizon; if NVDA rallies >30% from entry, take 50% profits; if NVDA falls >10% from entry within 30 days, add to increase to 4–6% portfolio weight. Hedge immediately by purchasing 3-month puts ~10–15% OTM sized to protect 1–2% portfolio value.
• Allocate 1–2% to GOOGL optional upside via 9–18 month LEAP call purchases (35–50% OTM) to capture TPU monetization optionality; increase to 2–3% if Meta/other hyperscalers announce cumulative TPU contracts ≥$1bn within 90 days.
• Enter a pair trade: Long GOOGL 1.5% / Short META 1.5% for 6–12 months to express TPU upside vs Meta capex uncertainty; close the pair if the spread tightens by 50% or after 12 months. Rebalance if Meta announces a strategy pivot (hardware divestiture or >$2bn TPU buy).
• Trigger-based reallocation: Monitor procurement/benchmark announcements over next 30–90 days; if a multi-year TPU deal ≥$2bn is confirmed, rotate 1–3% from NVDA into GOOGL within 30 days. If no material TPU procurement/benchmarks appear in 6 months, trim GOOGL option exposure by 50% and reallocate to cloud/software (MSFT, AMZN) or back to NVDA.