Nvidia to provide 1 gigawatt of AI chips, make 'significant investment' in OpenAI rival Thinking Machines Labs

Nvidia will supply upwards of 1 gigawatt of next-generation Vera Rubin AI chips to Thinking Machines Labs and is making a "significant investment" in the startup (amount undisclosed), with processors to be deployed early next year. Nvidia reported Q3 EPS $1.30 on revenue $57.01B, data center sales $51.2B (vs $49.3B est.), and gave Q4 revenue guidance of $65B ±2% (Wall Street $62B). The deal accelerates capacity for an OpenAI rival but raises circular-investing / competition concerns as chipmakers invest in startups that then buy their processors.

Analysis

Strategic minority investments by dominant infrastructure suppliers create a feedback loop that materially raises switching costs for enterprise AI buyers — not just via raw throughput but by locking in software, orchestration, and procurement relationships. Over 6–24 months this dynamic favors vendors that can sell a full-stack solution; competitors selling only chips or point components will need to match ecosystem depth (software, optics, power/cooling partners) or accept margin erosion.

On the supply chain front, demand pooling around integrated stacks amplifies pressure on adjacent suppliers (optical modules, high-voltage PDUs, liquid-cooling OEMs) and shortens their lead times, creating a window for outsized revenue growth for niche suppliers with available capacity. Conversely, foundry and advanced-assembly bottlenecks become a choke point: if capacity cannot scale as bookings front-load, deployments will be delayed and customers may seek alternative architectures or software-level optimization to reduce capacity needs.

Key risks that could flip the trade are regulatory scrutiny of “circular investing” and a LLM spending rebase. Both are binary and operate on different horizons — antitrust headlines can move multiples within days, while enterprise procurement repricing and capacity normalization play out over quarters. Practical catalysts to watch: public disclosures of investment terms, optics/power supplier orderbooks, and any enterprise capex guidance that shifts from linear growth to pull-forward or pull-back patterns.

Key Decisions for Investors

• Long NVDA via a 6–12 month call spread (buy calls, sell higher-strike calls) sized to 3–5% of portfolio; target 2:1 upside/downside payoff. Rationale: capture continued premium for integrated stacks while limiting theta. Hedge with a small (10% notional) out-of-the-money put to protect vs anti-trust headline risk.
• Long optics suppliers COHR and LITE (equal weight) for 6–12 months — buy outright or buy-to-open 9–12 month calls. Expect lead-time driven revenue beats; set stop at 20% downside or take profits on 30–50% appreciation. Risk: cyclical book-to-bill reversal if hyperscalers pause orders.
• Pair trade: long NVDA / short AMD, 6–12 month horizon, position size 1–2% net market-neutral. Rationale: asymmetric capture of ecosystem lock-in; if NVDA’s stack premium persists, AMD multiple compresses. Risk: AMD product wins or aggressive pricing could invert outcome; cap loss at 25% of pair notional.
• Event hedge: buy 1–3 month NVDA protective puts or a put spread before major regulatory filings or earnings/guidance windows to limit tail downside from a negative disclosure. Cost should be treated as insurance (budget ~25–40 bps of portfolio) given headline-driven volatility.