CognitoRiverDelta Patreon
news through November 9, 2026
← All reports
Category rollup

Public markets & the AI trade - 2026-W46

Week of November 9, 2026 · 18 min read Download PDF Share on X

Public markets & the AI trade · week 2026-W46: Nov 03 - Nov 09, 2026 · 1 subtopic(s) covered · 2823 words · expanded

Overview

The central narrative of this week is the violent transition of the AI trade from a software-centric "intelligence" play into a massive, high-stakes "physical infrastructure" arms race. For much of 2024 and 2025, the market focused on model performance and token efficiency; this week, the focus has pivoted decisively toward the hard realities of compute power, electricity, and the physical hardware required to host autonomous agents. This shift is best exemplified by the diverging paths of the "model makers" and the "infrastructure titans." While foundational model companies like Anthropic and OpenAI face staggering compute liabilities and high-stakes valuation milestones, companies like SpaceX are aggressively positioning themselves as the next generation of vertically integrated hyperscalers, leveraging both terrestrial data centers and the promise of orbital compute to bypass the growing bottlenecks of the terrestrial power grid.

This theme of "physicality" creates a palpable tension across all major tech signals. The massive capital expenditures required for AI are no longer just a line item on a balance sheet; they are becoming geopolitical and topographical issues, as seen in the political friction surrounding data center expansion in Texas following Democratic election wins. The regulatory hurdles facing Tesla’s autonomy in Japan and Australia further complicate the path for physical AI deployment. Furthermore, the potential convergence of the Tesla and SpaceX ecosystems suggests a future where AI value is not derived from the ability to "chat," but from the ability to move, sense, and act in the physical world—via Robotaxis, humanoid robots, and satellite-linked 5G networks. This week, the market is effectively re-pricing the entire sector, moving away from the "virtual" intelligence of the LLM era and toward the "physical AI" era of robotics and orbital infrastructure.

AI/tech investing signals

The most profound signal of the week is the sheer, aggressive scaling of AI compute infrastructure, which is moving from a stage of preparation into a stage of massive, high-risk deployment. This is most visible in the activities of SpaceX, which is rapidly becoming a dominant player in the compute landscape. The deployment of the Gunrange cluster (220,000 GB300 chips) in November and the SPAM cluster (200,000 inference chips) in December signals a move to capture the massive demand for both training and inference.

This expansion is not merely about adding capacity; it is about strategic vertical integration. SpaceX is attempting to control the entire stack: the chips, the power, the connectivity (via the Qualcomm 5G partnership to deploy a terrestrial network by 2027), and eventually the "hosting environment" itself. The reported expansion of SpaceX's AI revenue—increasing by a staggering 1,800% in Q3—underscores that this is already a core driver of the company's massive revenue growth. However, this scale brings immense complexity and potential risk. The proposed 6-gigawatt compute complex in Arizona, featuring up to nine 500,000 sq ft halls, illustrates the staggering physical footprint required. This terrestrial scaling, however, faces a "physical ceiling" in the form of grid constraints, long queues for power, and local political opposition, particularly in Texas where data center expansion faces uncertainty.

In response to these terrestrial bottlenecks, a new and highly speculative strategic pivot is emerging: "orbital compute." SpaceX's "Starmind" concept—moving data centers into space—is being framed as a hedge against the political and energetic limitations of Earth. While some analysts, such as Joseph Farah, argue that the laws of thermodynamics make space-based AI compute impractical, others like Jo Bhakdi suggest that the plummeting costs of Starship launches (with Raptor 3 engines costing <$1M compared to the ~$8M BE-4) could make orbital compute cheaper than terrestrial options by 2028. This debate highlights the central tension of the week: whether the AI trade will be choked by the limits of our terrestrial infrastructure or if it will leapfrog those limits through radical new deployment methods.

Parallel to this infrastructure build-out is a massive shift in the valuation models for the major players, particularly Tesla and SpaceX. Tesla's valuation is increasingly viewed not as a car company, but as a bet on "physical AI"—the scaling of Optimus humanoid robots (with a target of 11 million by 2030) and the Robotaxi fleet. This is reflected in the massive divergence in price targets, with ARK Invest projecting a $4,400 stock price by 2031 based on the sale of millions of robots and autonomous vehicles, whereas Wall Street maintains a conservative average target of ~$410. The expansion of Robotaxis into Fort Worth, making it the seventh driverless city, serves as a key execution metric for this thesis.

However, this "physical AI" thesis is being tested by market volatility. Following the midterms, both Tesla and SpaceX saw a combined market cap loss of over $420 billion, highlighting the sensitivity of these high-multiple stocks to political and macroeconomic shifts. Furthermore, the potential for a Tesla/SpaceX merger adds a layer of complexity to investor strategy. While a merger could provide a valuation "floor" through a shareholder premium, it also introduces an "overhang" if the transaction involves trading high-multiple Tesla stock for lower-multiple SpaceX stock. Analysts are divided on the timeline, with some predicting a 90%+ probability of a merger within 18 months, while others suggest a window of up to five years.

The "model maker" layer of the AI trade is facing its own reckoning. The imminent IPO of Anthropic is a pivotal moment that could "make or break" the current AI sentiment. The scale of commitment required by these companies is unprecedented; for instance, reports of Anthropic's $895 billion compute commitment and its reported $3 billion monthly compute costs suggest a level of capital intensity that could challenge long-term profitability. There is a growing concern that these companies are becoming "tenants" of the infrastructure providers. SpaceX's reported use of 90-day termination clauses in compute-leasing agreements gives them significant leverage over model companies, essentially allowing them to reprice services at spot rates and exert control over the very companies that rely on their hardware.

Finally, the hardware supply chain itself is experiencing significant tremors. The leadership scandal involving Nvidia's CEO, Jensen Huang, and the subsequent 5.4% stock drop following reports of a board investigation into his relationship with a suspected Chinese spy, serves as a reminder of the concentration risk inherent in the AI trade. If the "brain" of the AI hardware industry is disrupted, the entire ecosystem's momentum could stall. This volatility is compounded by the reality that the demand for AI compute is shifting toward "agentic" workflows—autonomous agents that require exponentially more tokens and reasoning power than human-prompted LLMs. This "token factory squeeze" favors those who own the vertical stack of power, chips, and connectivity, rather than those who simply manage the software layer.

Cross-cutting themes

The primary tension connecting all these developments is the struggle between virtual scalability and physical limits. The AI trade has long been driven by the idea that software scales almost infinitely with the addition of more parameters and more data. However, this week's developments show that this "virtual" expansion is crashing into "physical" walls: power grid capacity, semiconductor availability, political resistance to data centers (notably in Texas), and the thermodynamic challenges of space-based compute. As the industry moves from simple LLMs to autonomous agents, the demand for compute is not just growing linearly; it is hitting a structural bottleneck where the physical ability to provide power and cooling may dictate the pace of AI intelligence.

A second theme is the shift from "Intelligence as a Service" to "Agency as a Service." The transition from human-prompted LLMs to autonomous, agentic workflows is driving a massive increase in token demand, which in turn accelerates the need for the massive infrastructure projects seen in SpaceX and the high-stakes fundraising seen in OpenAI (reportedly raising $11 billion at a $1.25 trillion valuation). This shift makes the "physicality" of the trade even more critical; an agentic economy requires not just more compute, but more reliable and low-latency compute, driving the interest in 5G/satellite integration and orbital data centers to ensure agents can act in real-time without terrestrial latency or power interruptions.

Finally, there is the theme of Vertical Integration vs. Horizontal Fragmentation. The market is increasingly favoring the "Musk model"—the idea that to win the AI race, you must own the energy (Megapacks), the transport (Tesla/SpaceX), the satellite network (Starlink), and the robot (Optimus). This stands in stark contrast to the "Foundational Model" model (Anthropic, OpenAI), where companies are essentially high-stakes tenants renting the "physicality" from others. This creates a structural tension in the market: will the long-term winners be the ones who invent the smartest models, or the ones who own the ground (and the orbit) they run on? As Steven Mark Ryan notes, the "compute hyperscalers" hold a strategic advantage because they control the underlying substrate of the new economy.

Where sources agree

  • The Infrastructure Bottleneck: There is a broad consensus among analysts that the primary constraint on AI growth is no longer just software capability, but the availability of physical infrastructure—specifically power, chips, and data center space. The grid queues and political opposition to data centers are recognized as existential threats to scaling.
  • Insatiable Demand: Analysts across the board agree that demand for AI compute and memory is "insatiable." This is evidenced by the massive scale of SpaceX’s cluster deployments and AMD’s doubling of data center revenue.
  • The Autonomy Pivot: There is general agreement that Tesla's valuation is increasingly decoupled from its automotive business and is instead a speculative bet on the future of autonomy, specifically through the scaling of the Robotaxi fleet and the Optimus humanoid robot.
  • Agentic Demand: Sources agree that the move toward autonomous AI agents will exponentially increase the demand for "tokens" and reasoning tasks, fundamentally changing the scale of required compute and favoring those with vertically integrated stacks.

Where sources disagree

  • SpaceX Valuation and Revenue: Opinions are wildly divided. Some analysts, like Larry Goldberg and Jo Bhakdi, argue the company is "massively undervalued" due to its orbital potential, while others, like Nick Gibbs, believe its $2 trillion valuation is already properly priced. Furthermore, revenue projections range from Morgan Stanley's $175B for 2027 to Elon Musk's $1T ARR target for 2028.
  • The Feasibility of Orbital Compute: There is a fundamental scientific and economic disagreement. Joseph Farah argues that the laws of thermodynamics make space-based AI compute impractical, while Jo Bhakdi contends that the plummeting costs of Starship launches will make orbital compute more efficient than terrestrial options by 2028.
  • Tesla's Future Valuation: The gap between traditional Wall Street targets (~$410) and ARK Invest's hyper-bullish projection ($4,400 by 2031) represents a massive disagreement on the economic viability of scaling millions of Optimus robots and Robotaxis.
  • SpaceX 5G Timeline: A significant discrepancy exists between Elon Musk’s 2-3 year timeline for a terrestrial 5G rollout and the 8-year timeline projected by industry experts and Verizon's CEO.
  • Nvidia's Leadership Stability: Following the reports of a board investigation into Jensen Huang, views are split between those seeing a "screaming sell" catalyst and those believing national security interests will ensure his retention.

Numbers and claims to verify

  • $450B+ SpaceX revenue: A speculative projection from Jo Bhakdi based on 10GW of capacity.
  • $895B Anthropic compute commitment: An extreme scale figure cited by Prof G that requires verification against S1 filings.
  • $3 billion monthly compute cost for Anthropic: A high-impact claim from Steven Mark Ryan regarding the financial strain on model makers.
  • $1 trillion 2028 revenue target for SpaceX: A target set by Elon Musk that represents a massive deviation from previous $200B consensus models.
  • Tesla fleet count in Texas: Reports are wildly contradictory, ranging from 187 (BigGo) to 1,044 (Bacti) and 1,244 (Shaw).
  • $14B revenue in SpaceX's first profitable quarter: A claim from Randy Kirk that requires S1 verification.
  • 90% merger probability (Tesla/SpaceX) within 18 months: A highly speculative timeline provided by an ARK panel.
  • $225 billion cumulative AI spend by 2031: An ARK projection that requires verification against Tesla's cash flow.

Investment and strategic implications

  • The "Hyperscaler" Advantage: In a market defined by "token demand" and "power constraints," companies that control the physical stack (compute, power, connectivity) hold a structural strategic advantage. The "tenant" model—where foundational model companies lease compute—is inherently precarious, especially when infrastructure providers like SpaceX employ 90-day termination clauses that allow them to reprice services at spot rates.
  • Credit Risk in AI Financing: The massive capital requirements for AI infrastructure are shifting the financing model from free cash flow toward debt markets. This is creating significant credit risk; for example, Coreweave's risk levels are reportedly 20x higher than investment-grade credit. Investors should be wary of the "stalling AI spending" risk that could leave these heavily levered infrastructure plays vulnerable.
  • Sovereign AI as a Strategic Trend: Enterprises are moving toward "sovereign AI"—the desire to protect proprietary IP by owning or controlling their own compute and models rather than just reselling tokens from third-party providers. This trend is driven by the need for data security and long-term cost predictability.
  • The Memory Opportunity: As the AI buildout continues, long-term memory shortages extending into the 2030s may present a structural opportunity for players like Micron, as agentic workflows and massive datasets place unprecedented pressure on memory capacity.
  • Diversification of Compute Locality: Investors should monitor the shift from terrestrial to orbital compute. While scientifically debated, orbital compute represents a potential hedge against terrestrial energy shortages, grid failures, and the political "NIMBYism" that threatens data center expansion on the ground.

What to watch next week

  • Anthropic IPO Valuation: Watch for the setting of the valuation range; this will be a critical bellwether for the entire AI sector's appetite for risk and the perceived sustainability of high-compute-cost business models.
  • Nvidia's Official Response: Any formal statement regarding the board investigation into Jensen Huang will be a major volatility catalyst for the entire semiconductor sector.
  • Tesla Regulatory Developments: Monitor updates from Japan regarding revised safety standards for steering-wheel-free vehicles and developments in Australia regarding the advancement of AV laws to 2027.
  • SpaceX Revenue/Profitability Data: Look for any S1 filings or verified reports that can confirm the explosive growth in SpaceX's AI revenue and whether the $14B profitable quarter claim holds up under scrutiny.

Appendix: Individual perspectives

We see further because we stand on the shoulders of giants. The people in this appendix are some of those giants: their public work, followed and studied every day and every week, is how CognitoRiverDelta brings the world's AI knowledge and concerns together, so that, in time, our expert panel can help shape real solutions and paths forward for all of us. We learn by observing them, and we think of them as kindred spirits: people helping humanity and AI educate each other and contribute to solutions. Together they form a much larger community of minds, a worldwide panel; ours is a small copy of it. Most people don't want to become expert kayakers, reading rocks and rapids every day; yet that is how this fast-flowing river of change feels to many of us, and the evidence that could ease those fears, the grounded ones and the imagined ones, is still being gathered. That is the work. We credit their ideas here with gratitude; they are not affiliated with us and do not endorse this report.

SpaceX & The Infrastructure Race

  • Jo Bhakdi: Views SpaceX as a resilient, high-conviction holding. He argues that plummeting Starship launch costs will make orbital compute more efficient than terrestrial options by 2028 and believes economic incentives will protect data centers from political opposition.
  • Randy Kirk: Emphasizes the massive scale of SpaceX's compute expansion, noting the potential for $400B+ in revenue from 10GW of capacity. He views SpaceX as a dominant provider of both terrestrial and potentially orbital compute, driven by engine cost-efficiency (Raptor 3).
  • Larry Goldberg: Attributes 80% to 90% of SpaceX's valuation to its orbital data center potential and notes the company has secured a massive $40 billion NVIDIA AI chip financing package.
  • Steven Mark Ryan: Warns of the political risks to terrestrial data centers (specifically in Texas) and notes that SpaceX's ability to reprice compute via 90-day clauses gives it immense leverage over model companies.
  • Joseph Farah: Maintains a skeptical view of orbital compute, citing thermodynamic constraints as a fundamental barrier to space-based AI.

Tesla & Autonomy

  • Nick Gibbs: Argues that Tesla possesses a massive proprietary moat in autonomy due to its scaling robotaxi operations and believes SpaceX's $2 trillion valuation is properly priced pending proven profitability.
  • Herbert Ong: Tracks Tesla via execution metrics and the "valuation divergence" caused by the convergence of AI, compute, and space infrastructure.
  • Jo Bhakdi: Suggests that the inflection point for a Tesla re-pricing will be the deployment of 1,800–2,000 Cybercabs, which could trigger a ~5% earnings increase.

LLMs & AI Models

  • Prof G: Highlights the extreme scale of compute commitments (e.g., Anthropic's $895B) and the massive credit risks associated with AI infrastructure financing.
  • Andrej Karpathy: Sees the frontier of value shifting from parameter scaling to the efficiency of reasoning via reinforcement learning and the integration of agentic, tool-using workflows.
  • Scott Galloway: Expresses skepticism regarding the fundamental strength of high-valuation AI companies, citing the unsustainable cost of serving models like Grok relative to their revenue.
  • Alex Karp: Warns that AI vendors are overselling capabilities and risks to client sovereignty, viewing the current market as a "lag period" of implementation rather than a bubble.
  • Sarah Guo: Anticipates an "SaaS apocalypse" for niche point solutions, where core systems of record will be the only ones to survive the AI transition.

Macro & Social Impact

  • Jonathan Haidt: Signals high regulatory and social risk for AI in the education and youth sectors, noting a potential disruption to cognitive development and human attachment.
  • Peter Diamandis: Focuses on the necessity of power and chips for national "sovereign AI," arguing that these resources provide a near-certainty (90-99%) of achieving AI dominance.

Sources

Informational analysis synthesized by AI from sourced, dated material, curated by a human. Treat specific claims as unverified until checked. Not financial advice.

About · How this is made · Corrections

💬 Tell us