$SPCX profit!; $TSLA merger?; $NFLX sinks; Elice rebound; Anthropic S1

DavidDTawil avatar
David D. Tawil
@DavidDTawilOct 22, 2026

The Spaces dissected a packed macro and tech tape: a brief market recap gave way to deep dives on AI infrastructure demand, Anthropic’s S‑1 snippets, Nvidia’s strategic positioning, SpaceX’s blowout quarter, and a prospective Tesla–SpaceX merger. Brandon outlined Anthropic’s rapid revenue ramp, enormous cloud commitments, and mostly non‑cash losses; David and Eric debated whether hybrid local/cloud AI and AMD/Intel advances could pressure Nvidia over time. Carlo detailed accelerating on‑chain finance: tokenized equities, JPM Coin on Ethereum, and Visa’s stablecoin rails. A spirited panel clash weighed insatiable data center demand against tightening credit, power and materials bottlenecks, and contract risks for non‑hyperscalers. SpaceX’s revenue grew 160% YoY with Starlink cash flow subsidizing growth, while AI compute deals became a top line driver; special committees will evaluate a Tesla–SpaceX combination. Anas unpacked diesel geopolitics (limited real supply impact), and Yan quantified a 50x surge in LLM token usage and looming power/cooling constraints as Nvidia’s 600kW racks arrive. The session closed with implications for labor displacement, infrastructure supply chains, and near‑term earnings (Meta, Tesla) and power exposure.

Market Wrap, Agenda, and Cast

  • Host: David (moderator)
  • Core contributors: Brandon (AI/Anthropic, cybersecurity), Eric (compute strategy and semis), Paul (credit markets), Carlo (crypto/on-chain), Dave Levinson (macro/rates), Eladio (SpaceX/Tesla), Justin (SpaceX numbers), Anthony Napolaltano (data center investor/operator), Nikowski (sector rotation/technicals), Honest (energy policy/markets), Yan (AI infrastructure/engineering leader)
  • Agenda: SpaceX/Tesla, Anthropic S‑1 takeaways, Nvidia positioning, crypto tokenization/stablecoins, macro/liquidity/rates, sector rotations, energy geopolitics, and near-term earnings (Meta, Tesla, IBM, LRCX, TXN, etc.).

Key Market Stats (prior day/this morning, per David)

  • Indices: Dow +1.0%, S&P 500 +0.25%, Nasdaq flat; Russell 2000 +>1.5%.
  • Vol: VIX fell. DXY below 102.
  • Rates: 10Y ~5.12%, 30Y ~5.42%.
  • Commodities: Brent near ~$110, WTI ~$98. (Note: spot vs. futures dynamics highlighted later by Honest.)
  • Notable moves: Equifax (raised EBITDA; still −20% YTD), Hasbro +11% (hit from a strong game), 3M +5% (beat/raise; data center tie-up with Amazon), Northrop Grumman +4%, GM −~7% (20B buyback ignored; guidance/input costs punished), Meta −1.6% (fifth straight down day; results after close).

Anthropic S‑1 (WSJ leak) and the Capex/Semiconductor Arc

What’s in the S‑1 snippets (Brandon; David’s adds)

  • Revenue ramp: H1’2026 ~$16.3B vs. H1’2025 ~$1.1B (David: a 16× YoY ramp; phenomenal but unlikely to 16× again from here).
  • Loss profile: Net loss ~$80.7B, of which ~$72B is non‑cash (revaluation/accounting of early investments); operating loss implied ~8.7B. For an expected $1T+ valuation, proximity to profitability is notable (David).
  • Compute/cloud commitments: ~$9B compute spend in the first half; ~$450B of non‑cancellable cloud commitments across 7–10 years; future infra spend cited at $600–$750B (David).
  • GPU sourcing: Renting up to ~$86B of Nvidia chips.
  • Risk section: expected to be extensive (Brandon).

Nvidia positioning and the “picks and shovels” logic

  • PR/Global presence: Jensen Huang’s deliberate Europe-heavy tour (Paris, London, Stockholm, Berlin) exemplifies elite PR; he “shows up,” compounding leadership (David).
  • Enterprise choices in flux: Frontier vs. open-source vs. custom models; on‑prem vs. cloud. Regardless, Nvidia remains the picks-and-shovels winner across stacks (David).
  • Diversification: Nvidia spreads risk across many partners (CoreWeave and others); avoids concentration risk (David). Potential recap needs at certain partners noted, but Nvidia not “all in” on one basket.
  • Local AI trend: Nvidia accelerating tooling for local compute while keeping hyperscaler and frontier lab ties (Brandon).
  • Competition and margins: Eric argues AMD/Intel will narrow the performance gap (shared TSMC nodes, AMD’s deep TSMC relationship), forcing Nvidia to price down over time; margins compress → potential systemic risks if Nvidia doesn’t pre‑fund/backstop ecosystem scale (Eric). Paul counters that access to process tech ≠ guaranteed wafer capacity/priority; TSMC’s allocation/customer service matters materially; capacity, not design kits, is the constraint (Paul).

Security PSA (Brandon)

  • Major vulnerabilities emerged in stacks underpinning many sites (e.g., PHP/Node.js; WordPress risk). Patches imminent; teams should assess exposure and patch promptly.

SpaceX: Blowout Quarter, AI Compute, and Prospective Tesla Combination

Reported results and guidance (Eladio; David; Justin)

  • Beat: SpaceX topped quarterly revenue estimates by ~$2B.
  • Growth: ~160% YoY revenue growth (David).
  • Profitability: Net income ~$4.4B vs. ~$1.1B loss last year (Justin).
  • Guidance: FY revenue guide ~$33–37B vs. a prior ~$22B context; commentary referenced “$100B December annualized” aspiration (speaker claim) and 2.5–3.0 GW compute capacity (Eladio). Expect volatility from additional float: up to ~44% of shares available to trade post report timing and lockup mechanics (Eladio).
  • Mix and cash flow: Starlink is the cash cow with large margins and recurring cash flow; currently subsidizes broader initiatives (David). However, the single largest revenue line is AI/Cloud compute—and it’s set to outgrow other segments (David; Eladio). Justin: ~$7.5B from multiple cloud compute rental deals (not just a single lab), 800k paying Grok users (small revenue now; rising).
  • Strategic framing: SpaceX is not a “pure play”—investors get a bundle: Starlink, launch, AI compute, and more (David).

Tesla–SpaceX combination (Justin; David)

  • Elon confirmed on the call that special committees at both boards are evaluating a merger; Elon is recused from the process.
  • Independent advisors on both sides; needs majority (50%+) votes of both shareholder bases—non‑trivial (David, Justin).
  • Eladio’s thesis: Market tends to reward Elon’s timing; sees potential “No. 3 in S&P 500” scale after combination; argues strong mechanical index levers if added to the S&P 500 over time.

Notable ownership/context tidbit (Brandon; David)

  • Anthropic: Eric Schmidt allegedly owns ~1% vs. Dario Amodei ~2%; at $1T valuation, Schmidt could see ~$20B (Brandon, David). Brandon notes a personal‑history wrinkle among principals—illustrating how Silicon Valley networks operate.

Macro, Liquidity, and Credit: The Bull–Bear Debate

Dave Levinson’s caution

  • Yield curves “melting,” long duration ripping (ZROZ ETF up sharply in two weeks)—signaling a violent bid for duration. He views this as “terrible for tech” near-term because of widening credit spreads and a looming oxygen (credit) shortage.
  • Vulnerabilities: low-margin GPU renters (e.g., CoreWeave-like operators, “Iron”) face funding stress as CCC costs rise; expects 52-week lows and consolidation; policy arsenal that backstopped tech drawdowns is largely spent.
  • Asset price risk: If GPU rental rates and GPU prices fall, large write‑downs could hit, pressuring profitability. Amazon cited as having tens of billions of chips not yet fully deployed—a red flag for later price/rental normalization.
  • Structural view: Demand will explode, but capital availability (cost/credit) is the binding constraint; DC bill of materials will reprice lower as chips normalize (amortization drag on prior vintages); more CDS/credit-led pressure to come.

David’s counterpoints

  • U.S. tech’s current revenue/profit growth is real and, barring exogenous shocks, likely to persist 12–18 months; global capital re‑allocation favors U.S. tech when local markets/sov debt underperform.
  • Contracts and end demand: not riskless. He flags two under‑appreciated bear points: (1) corporate AI spend/utilization can pause (e.g., Deutsche Bank’s CEO signaling pullback to selected frontier usage); (2) “locked” contracts are negotiable—large customers can and do renegotiate or walk away to preserve relationships.

Operator/investor views from the field

  • Anthony Napolaltano: Demand for AI data centers “insatiable,” likely years of undersupply; co‑location (build, rent long‑term to hyperscalers) is a conservative model; bonds tied to hyperscaler‑backstopped projects trade firm. Acknowledges some cloud upstarts will fail, but believes we’re too early in the build‑out for systemic bust.
  • Nikowski: Sees pronounced sector rotation—E&C, industrials, and many suppliers into DC buildouts are deeply sold off and underperforming; high yield spreads rising; multiple compression spreading. Prefers tactical posture; avoid 35–50% drawdowns; rotate into oversold leaders in non‑tech when the tape favors it.
  • Paul: From a major credit conference: big allocators are actively buying secondaries that retail/interval funds are forced to sell (steep discounts). Emphasizes data centers are not “just chips”—permits, steel, concrete, and municipal approvals are real bottlenecks (steel scarcity already visible). Predicts heavier automation of analyst/associate work via “hallucination‑controlled” AI stacks; heavy compliance/oversight to stop data leaks.

Nvidia, China Orders, and Capacity

  • China demand: ByteDance, Alibaba, Tencent reportedly placed orders for 3M+ RTX Pro 5500s (~$35B order book), but Nvidia can only partially fulfill due to constraints (David). Policy angle: even with export limits, Chinese demand is strong for allowed SKUs—positive signal for Nvidia and U.S. policy leverage.

Crypto, Tokenization, and Stablecoins (Carlo)

  • Bitcoin ~83,000; broad on‑chain asset adoption accelerating.
  • Tokenized equities split: (1) OKX + NYSE seek SEC‑compliant tokenized stocks, but 16 issuers have objected; (2) “EZ A” (a blockchain/AI dev platform) seeks the opposite—tokenize its own shares day‑one via a direct listing (F‑1 filed), letting existing holders sell on‑chain, no new primary capital.
  • Banks and stables: JPM Coin now live on Ethereum for eligible institutions—fast on‑chain movement of deposits inside JPM’s ecosystem; Visa x Mesh to enable crypto (incl. stablecoins) spend at merchants via Visa rails—reducing last‑mile friction. Carlo’s thesis: money, markets, and assets are moving on‑chain; watch banks and networks formalize it.

Energy Update and Geopolitics (Honest)

  • Prices: Futures slipped ~1%, but spot remains elevated; current structure is economically destructive if sustained.
  • Diesel diplomacy before elections:
    • “Trump–Putin” diesel: Russia to ship diesel to the U.S. despite sanctions (workarounds via swaps are likely). First cargo equates to ~1 day of Northeast heating oil demand—economically trivial for the U.S., but a political optics win for Putin (deterrence message to Ukraine vis‑à‑vis refineries).
    • U.S. SPR: Offering up to ~300k bbl of heating oil (also ~1 day of NE consumption)—again economically small, politically symbolic.
    • China pre‑emptively halted diesel exports—offsetting flows; net market impact ~zero.
    • EU (France/Germany) signed diesel supply deals with UAE/Saudi to hedge against potential U.S. export bans; prioritize winter security over short‑term U.S. political requests.
  • East Med gas chessboard: TotalEnergies discussing a Cyprus-to‑Egypt pipeline to utilize underused Egyptian LNG plants; Turkey counters with a new maritime deal with Libya—classic energy geopolitics.
  • Policy symmetry: “Waivers to keep talks alive” mirrors Biden’s prior Iran policy—same playbook, different country (Honest).

Demand Tsunami From the Trenches: AI Tokens, Power Density, and ‘Brain Factories’ (Yan)

  • Observed usage explosion: OpenRouter volume grew ~50× in 12 months—from ~3.5T tokens (≈words) last October to ~160T. That’s mostly knowledge work: coding, tasks, workflows. Next wave is multimodal.
  • Multimodal leverage: Images consume ~60× tokens vs. text; video ~200×. Voice and analytics artifacts will be tokenized too. Net: data centers face a demand “tsunami,” with Chinese model inference also seeking U.S. compute given domestic constraints.
  • Enterprise adoption: Inside Yan’s 6,000‑engineer org, only ~10–15% deeply engaged today; usage is growing 5× in six months. Subscription under‑utilization is common—the “gym membership” effect—healthy for model providers’ unit economics; power‑users exist but are a minority.
  • Power/cooling limits: Current mainstream AI racks ~120–140 kW per 3×3 ft rack; Nvidia’s coming Rubin Ultra is ~600 kW per rack (5×). The question: which data centers have the space, power delivery, and cooling to swap in 600 kW racks at scale?
  • Space vs. ground compute: SpaceX’s orbital data center concepts must contend with the sheer practicality of delivering and cooling 600 kW‑class density; Yan expects ground‑based “brain factories” (his term) to dominate near‑term due to power/cooling/logistics advantages.

Sector Watch: Media and Late Entrants

  • Oracle / Paramount–Skydance complex: David flags leverage and performance concerns. Larry Ellison reportedly borrowed >$450M against holdings (Oracle, SpaceX, Tesla) to fund deals; one media equity (SKYD) fell from $12 to ~$7.15 (Laurene Powell Jobs down ~$100M on a disclosed purchase). Thesis: Oracle is highly levered and late to the AI party; media economics (esp. live sports) are brutally competitive with thin margins; Netflix’s quarter seen as weak with thin content pipeline and fierce sports bidding by rivals.

Additional Notables and Housekeeping

  • Korea listing: A “junior Korean company” to CoreWeave plunged ~32% on debut, then hit a +30% daily up‑limit next session—but remains below initial pricing (David; Levinson cautions base math still negative).
  • Power as investment theme: Consensus on stage—power is not getting cheaper; owning power generation/rights may be a prudent portfolio hedge (David).
  • Upcoming earnings: Meta and Tesla report after close today; IBM, Lam Research, Texas Instruments, Southwest, United Rentals also on deck.

Synthesis: Where There’s Agreement—and Where There Isn’t

  • Broad agreement:
    • AI demand growth is real and shifting rapidly to multimodal, multiplying compute needs.
    • Physical bottlenecks are binding: power, cooling, steel, permitting, and high‑density rack readiness.
    • Credit conditions matter; funding costs for lower‑rated operators are rising; selection will be crucial.
    • SpaceX delivered a blowout, with Starlink cash flow and AI compute emerging as twin engines.
    • Power economics could be a defining constraint; power exposure looks strategic.
  • Active debates:
    • Timing/magnitude of a GPU/rental price normalization and resultant write‑downs (Levinson cautious vs. Anthony/Eladio optimistic on sustained demand and hyperscaler credit quality).
    • Nvidia’s medium‑term margin resilience in the face of AMD/Intel competition and TSMC capacity allocation (Eric sees margin pressure; Paul stresses capacity priority and allocation as the chokepoint).
    • Equity leadership durability: David/Eladio emphasize ongoing US tech earnings strength over 12–18 months; Nikowski/Levinson point to spreading multiple compression, HY spread widening, and the need for tactical rotation.

Practical Takeaways

  • For operators/investors in AI infra:
    • Stress‑test power and cooling plans for 600 kW/rack era; evaluate facilities for swap/retrofit feasibility.
    • Prioritize counterparties with hyperscaler‑level credit and scrutinize contract terms for renegotiation risk.
    • Expect steel/construction/permit timelines to be gating factors; de‑risk supply chains.
  • For AI adopters/enterprises:
    • Anticipate seamless hybrid (local + cloud) usage models to optimize cost/performance; implement governance to prevent data leakage.
    • Recognize “gym membership” economics—avoid over‑buying tiers that your users won’t consume; monitor utilization.
  • For portfolio construction:
    • Acknowledge divergence inside “AI plays”: co‑lo REIT‑like vs. full‑stack cloud vs. GPU renters—distinct risk/return.
    • Consider power producers and grid‑adjacent assets as structural beneficiaries.
    • Remain tactical; respect credit spread signals and rotations while tracking genuine earnings strength.

Closing Note

  • Tomorrow’s focus: reactions to Meta/Tesla prints, plus IBM/LRCX/TXN and more. The room emphasized re‑sharing and timely presence to cover the deluge.