Most important: Tesla AI compute capex intensity and Dojo rampIn the print: capex level, energy storage GWh, auto gross margin ex-creditsOn the call: Dojo procurement, GPU cluster expansion, Optimus timelineDownstream: NVDA, AVGO, TSMC, SK Hynix
House view
Street prices Tesla as an auto/robotaxi story. We read this print as an AI compute capex signal, tracking whether Tesla's $20B+ 2026 capex guide pulls GPU and custom silicon capacity from hyperscaler procurement. Read-across to NVDA allocation and TSMC advanced packaging.
What is priced in
Street expects a soft auto quarter given 358K deliveries and $22.3B revenue. Not priced in: the split of $20B+ capex between AI compute (Dojo clusters, training GPUs) and factory buildouts. A large AI compute share would tighten NVDA and AVGO custom silicon allocation at the margin.
What to extract from this callRanked by constraint impact
Priority 1 · primary read
Tesla AI compute capex intensity and Dojo ramp
Tesla guided $20B+ 2026 capex across six factories plus AI. The AI compute share determines incremental GPU demand hitting NVDA allocation and custom silicon orders flowing to AVGO and TSMC advanced packaging.
✓Capex at or above $5B for Q1 with management citing rising AI training cluster and Dojo hardware spend, naming GPU or custom chip procurement volumes.
✗Capex below $3B with AI compute spending deferred to H2 or Dojo delays cited, easing near-term GPU and packaging demand on NVDA and TSMC.
Priority 2
Custom AI silicon procurement cadence (Dojo chip orders)
Dojo chips route through TSMC advanced nodes and AVGO packaging. Acceleration of Dojo deployment signals incremental custom silicon demand competing with hyperscaler orders at TSMC.
✓Management discloses expanded Dojo cluster deployments or new chip tape-outs, with specific node or volume references pointing to TSMC capacity pull.
✗Tesla pivots further toward off-the-shelf NVDA GPUs for training, reducing custom silicon demand at TSMC and AVGO while concentrating spend on NVDA.
Priority 3
Energy storage cell sourcing and battery supply tightness
Tesla deployed 8.8 GWh in Q1 with all storage cells sourced from China under $200M+/quarter tariff headwinds. Shifts in sourcing signal CATL and BYD capacity allocation changes affecting LFP supply for the broader storage market.
Working the eventRelease drop vs. Q&A
In the release · first 60 seconds
EPS
$0.37 consensus (Q1 2026)
Beat or miss vs $0.37 is secondary. Focus on whether opex or capex charges drove the delta.
Revenue
$22.32B consensus (Q1 2026)
Check energy segment revenue share. Energy was $12.8B full-year 2025. A rising Q1 mix signals storage demand pull.
Margin compression below 16% with 358K deliveries would signal pricing pressure absorbing tariff costs rather than passing through.
Energy storage deployed (GWh)
8.8 GWh (Q1 2026 production data)
Confirm revenue recognition on 8.8 GWh. Any upside signals Megapack backlog conversion accelerating.
Full-year capex guide reaffirmation
>$20B (guided Q4 2025)
Downstream readsOutcome → what it means for names we care about
Capex above $5B + Dojo ramp accelerated
NVDA and TSMC advanced packaging capacity tightens. AVGO custom silicon demand rises. SK Hynix high-bandwidth memory allocation stays constrained.
Capex above $5B + GPU-heavy, Dojo delayed
NVDA GPU allocation tightens as Tesla adds to hyperscaler demand. TSMC and AVGO custom silicon pressure eases near-term. Mixed for SK Hynix.
Capex below $3B + full-year guide maintained
Near-term GPU and packaging demand from Tesla eases for NVDA and TSMC. Constraint pressure shifts to H2, giving smaller buyers a Q2-Q3 window.
Capex below $3B + full-year guide cut below $20B
Tesla AI compute demand drops out of the constraint equation. NVDA and TSMC lose an incremental demand source. Eases custom silicon pressure at AVGO.
2074 signals · 43 high-qualityResearch read-through · not a trade recommendation
✓Energy storage deployments above 10 GWh with new non-China cell supply contracts announced, signaling diversified LFP procurement tightening alternative suppliers.
✗Deployments flat or down with tariff costs rising and no new cell sourcing, suggesting Tesla absorbs pain without pulling LFP capacity from competitors.
Any raise above $20B or specific AI compute allocation percentage is the key constraint signal for NVDA and TSMC.
On the call · where the read moves
Dojo chip deployment timeline and cluster scale
Accelerated Dojo ramp pulls TSMC advanced node capacity and AVGO packaging, tightening custom silicon supply for other customers.