Musk claims Tesla will restart work on its Dojo supercomputer

Elon Musk said Tesla will restart work on Dojo3, the company’s third‑generation in‑house supercomputer project previously disbanded while resources were shifted to onboard AI chips (AI5/AI6). Musk tied the revival to progress on the AI5 chip design and noted AI6 chips will be manufactured by Samsung in Texas under a reported $16 billion agreement; Dojo’s stated purpose is large‑scale video and sensor training for Full Self‑Driving, though Musk also described Dojo3 as an experimental space‑based compute concept and experts remain skeptical. The announcement signals a strategic re‑prioritization of training infrastructure that could matter for Tesla’s long‑term AI roadmap but is speculative and unlikely to change near‑term financials.

Analysis

Market structure: Restarting Dojo3 signals Tesla pursuing verticalized training capacity that could reduce its reliance on cloud/GPU providers and raise the long-term marginal value of its FSD data. Direct winners: TSLA (IP, lower per-sample training cost), SSNLF (Samsung fabs via $16bn deal) and, if space-based compute materializes, launch/satellite suppliers; losers: third‑party training GPU vendors (partial TAM risk) and hyperscalers for Tesla-specific workloads. Expect modest re‑allocation of compute spend over 12–36 months rather than abrupt market share shifts.

Risk assessment: Key tail risks include technical failure of space‑based cooling/latency economics, regulatory limits on uplink/downlink or vehicle data transfers, and foundry execution risk at Samsung; low‑probability but high‑impact outcomes could wipe out multibillion-dollar capex. Immediate reaction (days) will be sentiment volatility around Musk statements; 3–12 month horizon hinges on Samsung fab ramp and Tesla capex disclosures; 1–3 years determines whether Dojo materially reduces inference/training spend outside Tesla.

Trade implications: Tactical plays favor trading volatility in TSLA options around earnings and Samsung fab cadence: buy 60–90 day straddles ahead of each catalyst and consider 18–24 month LEAP calls on TSLA if you ascribe >30% upside from FSD monetization. Consider selective longs in SSNLF (exposure to AI6 production) sized 1–2% of portfolio and small, speculative exposure (0.25–0.5%) to RKLB/Rocket Lab if SpaceX/launch announcements accelerate.

Contrarian angle: The market is underestimating execution risk and cost of space compute — launch, maintenance, latency, and regulation likely keep Dojo3 experimental for 3+ years. If Tesla diverts engineering from AI5/AI6 to Dojo3, FSD timelines could slip, creating short opportunities on misses. Historical parallel: hyped vertical AI hardware projects frequently see multi‑quarter delays before commercial returns, so price in execution skepticism over next 6–12 months.

Key Decisions for Investors

• Establish a tactical 1–2% long position in TSLA using 18–24 month LEAP calls (target delta ~0.30) to capture upside from successful FSD + in‑house training; trim to half if TSLA rallies >30% within 6 months or after a Samsung fab miss is announced.
• Buy 60–90 day ATM straddles sized 0.5–1% NAV on TSLA 1–2 weeks before each quarterly earnings and before Samsung Texas fab ramp milestones to monetize expected spikes in implied volatility; close within 7–14 days after the catalyst.
• Allocate 1–2% long position to Samsung Electronics ADR (SSNLF) to play AI6 manufacturing upside; scale up to 3% only if Samsung publishes wafer‑starts or capacity buildout dates consistent with >50k wafer starts/year within 12 months.
• Initiate a small, speculative 0.25–0.5% long position in Rocket Lab (RKLB) or comparable launch supplier, conditional: increase to 1% only if Tesla/SpaceX/other public disclosure shows >100 dedicated launches or satellite orders tied to Dojo3 within 18 months; otherwise exit position at 6 months.
• If Tesla misses FSD milestones or delays AI5/AI6 delivery by >2 quarters, short 0.5–1% TSLA size (or buy puts) targeting a 20–40% downside over the following 3–9 months; set stop‑loss at 15% adverse move from entry.