Public markets & the AI trade - 2026-W40

Public markets & the AI trade · week 2026-W40: Sep 27 - Oct 03, 2026 · 1 subtopic(s) covered · 2254 words · expanded

Overview

The overarching storyline of this week is the profound, accelerating transition of the AI trade from a digital-first software phenomenon into a massive, capital-intensive "physicality" play. For much of 2023 and 2024, the market’s attention was concentrated on the "intelligence" layer—the Large Language Models (LLMs) and the software applications that run on them. However, the current cycle of developments suggests that the most significant investing signals are now shifting toward the physical infrastructure required to house, power, and deploy that intelligence. This shift is characterized by a convergence of high-end semiconductor diversification, the integration of AI into humanoid robotics, and the radical expansion of AI into orbital infrastructure.

The week’s disparate threads connect through a singular, massive cycle of capital expenditure. Whether it is Dan Ives’s observation of the multi-trillion-dollar CapEx pool or Randy Kirk’s focus on AI infrastructure leasing, the underlying message is that the "AI trade" is no longer just about who builds the smartest model, but who owns the physical hardware, the energy-efficient data centers, and the robotic interfaces that allow that intelligence to interact with the real world. We are witnessing a fundamental collision between the digital world of silicon and the physical world of heavy industry, space, and robotics. This transition marks the end of the "pure software" era of AI and the beginning of an era defined by physical deployment and industrial scaling.

AI/tech investing signals

The landscape of AI investing signals this week revealed a widening divergence between traditional software valuations and the staggering, often speculative, valuations being applied to companies that bridge AI with physical hardware and orbital infrastructure. The market is moving away from evaluating AI based on code efficiency and toward evaluating it based on physical throughput, energy management, and manufacturing scale.

The Semiconductor Milestone and Hardware Diversification A critical signal for the hardware layer was AMD reaching a $1 trillion market capitalization, marking a significant structural shift in the semiconductor ecosystem. With the stock increasing over 180% in 2026, this milestone suggests that the "AI hardware" signal is no longer a mono-culture dominated solely by Nvidia. For investors, this diversification is a key metric for the health of the entire AI trade; it indicates that the demand for AI compute is broad enough to support multiple dominant players and a more complex, multi-vendor semiconductor stack.

This milestone also provides context for the massive scale of the current investment cycle. Dan Ives notes that the projected $4–$5 trillion in AI CapEx has seen only about 15% of its total deployment. AMD’s rise suggests that as this capital flows, it will not be concentrated in a single point of failure but will distribute across a widening array of silicon providers. This diversification is essential for mitigating the supply-chain constraints—such as CoWoS packaging and HBM memory availability—that currently dictate the top-line delivery potential for all accelerator platforms.

The SpaceX-Starlink Valuation Tension Perhaps the most volatile and analytically complex signal this week came from the aerospace and space-compute sector, where there is a massive, unresolved tension regarding the valuation of SpaceX and its satellite internet arm, Starlink. There is a significant discrepancy in current reporting: TradingKey cites a SpaceX IPO debut in June at $135, valuing the company at $1.75 trillion, while other analyses, such as those from Steven Mark Ryan, suggest the total valuation is currently under $2 trillion.

However, the more profound analytical signal lies in the long-term projected re-rating of the company. Ron Baron anticipates that Starlink could reach $1 trillion in revenue within a decade, potentially driving a valuation toward $14–$15 trillion based on estimated EBITDA of $700–$800 billion. This astronomical figure is underpinned by a radical redefinition of SpaceX's core business. Elon Musk has predicted that within approximately five years, AI will account for 99% of SpaceX's total value.

This suggests that SpaceX is being re-rated by the market not as a launch company, but as an AI-driven infrastructure company. This aligns with Jo Bhakdi’s perspective that the economic dominance of space-based computing could eventually undercut terrestrial costs by 2035. For investors, the signal here is the potential for a "space-based compute" multiplier—where Starlink V3 bandwidth increases and Starship's capabilities combine to create an orbital intelligence layer that operates outside the terrestrial constraints of power grids and land-use regulations.

Robotics and the Scaling of Physical AI On the terrestrial front, the focus has shifted to the "last mile" of AI: the transition from digital reasoning to physical action. Tesla’s reported audit of Chinese suppliers to ramp up the Optimus Gen 3 robot represents a significant strategic signal. The target price point of $20,000 for a humanoid robot, as suggested by TrendForce, represents an attempt to move AI from a high-cost enterprise tool to a mass-market physical utility.

This move toward low-cost, large-scale manufacturing is a critical indicator of whether AI can break out of the data center and into the labor market. If Tesla can achieve these pricing goals, the "physical AI" trade moves from speculative software deployment to a massive industrial scaling event. This transition is supported by analysts like Randy Kirk, who argues that the velocity of physical factory deployment and the growth of AI infrastructure leasing revenue are the primary signals that the market is currently undervaluing players like Tesla and SpaceX.

Cross-cutting themes

The primary cross-cutting theme this week is the Physicalization of the AI Trade. We are observing a convergence where the digital, the orbital, and the terrestrial are all being unified by a single driver: massive capital expenditure on intelligence-capable hardware. The "trade" is no longer about the intelligence itself, but about the physical vessels (robots, satellites, data centers) that deliver it.

This leads to a second major theme: the Tension Between Digital Intelligence and Physical Constraints. While analysts like Dan Ives and Chamath Palihapitiya emphasize the massive growth potential and the role of AI in driving real GDP (noting that AI accounts for much of the 1.5% real GDP growth), the sector's "Durable Background" highlights looming non-silicon bottlenecks. These include power grid availability (manifested in regional interconnection queues), datacenter interconnection density, and the high cost of inference. The critical question for the market is whether the projected $4–$5 trillion in CapEx can overcome these physical constraints—power, cooling, and orbital bandwidth—fast enough to justify current valuations.

Finally, there is a clear theme of Growth Velocity vs. Economic Realism. There is a palpable divide between those focusing on the "multiplier effect" of spending (Ives's $5-$6 multiplier) and those focusing on the "ROI gap" and unit economics (Wissner-Gross). As the market moves from the "building" phase to the "monetization" phase, the focus is shifting toward whether software vendors can drive net revenue retention (NRR) and whether they can prevent gross margin dilution caused by the high cost of inference.

Where sources agree

Despite the volatility in specific valuations, there is a broad consensus across the analyzed perspectives regarding the structural direction of the market. First, there is agreement on The Scale of the AI Trade: analysts like Ives, Palihapitiya, and Kirk view AI not as a niche sub-sector, but as a fundamental driver of the broader macroeconomy and US GDP growth. Second, there is a shared recognition of The Importance of Hardware/Infrastructure: there is implicit agreement that the current phase of the AI cycle is defined by massive capital expenditures in physical assets, ranging from semiconductor capacity to satellite constellations. Finally, sources across the board—from the discussions on Tesla robotics to SpaceX's AI integration—point toward a future where AI's primary value is realized through Physical Deployment, moving the intelligence from the screen into the real world via robots, space-based compute, and automated factories.

Where sources disagree

The sources diverge most sharply on the specifics of valuation and the pacing of implementation.

• SpaceX Valuations: There is a massive discrepancy between current reported IPO data ($1.75 trillion) and the long-term projected valuations of Starlink ($14–$15 trillion). This creates a tension between those viewing SpaceX as a high-value private aerospace company and those viewing it as a future trillion-dollar AI-infrastructure behemoth.
• AI Implementation Pacing: A fundamental disagreement exists between the "accelerated deployment" camp (Ives, Palihapitiya), who see massive under-deployment and a multi-year bull market, and the "safety/cautionary" camp, led by Sam Altman, who prioritizes managing exponential risks and rejects rapid commercial or IPO milestones that could compromise catastrophic risk management.
• The Driver of AI Value: There is an ongoing debate over whether AI value is ultimately driven by software revenue execution and free cash flow (Wissner-Gross) or by the ownership of physical infrastructure, proprietary datasets, and the deployment of physical assets like Cybercabs and Starlink (Diamandis, Bhakdi, Kirk).
• The Economic Reality of AI: While some see AI as the primary engine of GDP (Palihapitiya), others (Wissner-Gross) argue that the market is currently masking fundamental challenges, noting that many AI companies are still struggling with revenue execution and that the delay in companies like OpenAI's IPO may be tied to these economic realities.

Numbers and claims to verify

• SpaceX IPO Data: The claim of a $135 debut price and a $1.75 trillion valuation on June 12 (TradingKey) requires verification, as it conflicts with the established understanding of SpaceX's private status and current valuation estimates (<$2 trillion).
• Starlink's $14–$15 Trillion Valuation: This is a highly speculative long-term forecast from Ron Baron. Its validity depends entirely on the massive EBITDA targets ($700–$800 billion) and the assumption of Starlink V3 bandwidth dominance.
• Tesla's Optimus Pricing: The $20,000 price point for Optimus Gen 3 (TrendForce) is a stated target for mass-market utility, not a confirmed production reality.
• AMD's 2026 Performance: The report of AMD stock being up over 180% in 2026 (CNBC Tech) must be cross-referenced with actual market performance for the current period.
• SpaceX's AI Value Contribution: Elon Musk’s prediction that AI will account for 99% of SpaceX's value in five years is a predictive assertion that lacks a verifiable baseline.

Investment and strategic implications

• Semiconductor Diversification: The rise of AMD to a $1 trillion market cap is a signal that the hardware trade is broadening. Investors should look beyond the primary provider (Nvidia) toward the secondary tier of the semiconductor supply chain, especially those companies addressing CoWoS packaging and HBM memory needs, as the $4-5 trillion CapEx cycle distributes across the ecosystem.
• The "Physical AI" Pivot: Strategic value is shifting toward companies that can bridge the gap between digital intelligence and physical utility. This includes Tesla (robotics and autonomous transport) and SpaceX (orbital compute). The "winners" of the next phase will likely be those who can scale physical deployment at low unit costs.
• Infrastructure vs. Software Risk: A critical distinction is emerging between "CapEx winners"—those providing the power, chips, and space-based compute—and "Software earners." Investors must scrutinize software companies for gross margin dilution; if inference costs are not managed, the traditional 75-85% SaaS margins could slide into the 60-70% range, fundamentally altering their valuation models.
• High-Alpha Space Exposure: The massive gap between current SpaceX valuations and the projected Starlink valuation suggests that space-based infrastructure represents a high-risk, high-reward frontier. This is a bet on the convergence of Starship's launch capacity and the massive bandwidth increases of Starlink V3.
• Biological Convergence: As highlighted by Peter Diamandis, the intersection of AI and synthetic biology—specifically the use of proprietary biological datasets for phenotype design—represents a separate but significant frontier for AI-driven value creation.

What to watch next week

• Hyperscaler CapEx Guidance: Any updates from Microsoft, Alphabet, Amazon, or Meta regarding their CapEx-to-revenue ratios will be the most significant leading indicator for the semiconductor and hardware sectors.
• Tesla Supply Chain Developments: Further reports on Tesla's audits of Chinese suppliers for Optimus Gen 3 will be essential to determining the timeline for physical AI scaling.
• Software Margin Integrity: Watch for any industry disclosures regarding the impact of AI inference costs on software gross margins, which will serve as a reality check on the software monetization thesis.
• Physical Bottleneck Indicators: Monitoring updates on CoWoS packaging capacity and HBM contract pricing will be crucial to determining if the hardware growth is demand-constrained or supply-constrained.
• SpaceX Leasing Metrics: Watch for any data regarding SpaceX's projected AI leasing revenue, which is a key metric for analysts like Randy Kirk to determine if the company is undervalued.

Appendix: Individual perspectives

• Chamath Palihapitiya: Argues the US economy is fundamentally dependent on the AI trade and Big Tech CapEx, asserting that AI drives nearly all real GDP growth.
• Dan Ives: Maintains an aggressively bullish outlook, citing massive under-deployed capital spending and a significant ecosystem multiplier of $5-$6 for every dollar spent.
• Dr. Alex Wissner-Gross: Focuses on fundamental economic metrics and unit economics; notes that OpenAI's IPO delay is tied to revenue execution challenges and highlights Anthropic's superior free cash flow performance.
• Jo Bhakdi: Views Tesla as a leader in physical AI and autonomous transport; projects that 1,800 deployed Cybercabs would trigger a cash flow inflection and that space-based compute will undercut terrestrial costs by 2035.
• Peter Diamandis: Sees the convergence of AI and synthetic biology as a major signal, driven by proprietary biological datasets and AI-enabled biological programming.
• Randy Kirk: Uses the velocity of physical factory deployment and the growth of AI infrastructure leasing revenue as primary signals for the undervaluation of Tesla and SpaceX.
• Sam Altman: Prioritizes safety-driven development pacing over immediate commercial milestones, rejecting a 2026 IPO as ill-advised due to the necessity of managing catastrophic AI risks.

Sources

• (News) Tesla Reportedly Audits China Suppliers for Optimus Gen 3 Ramp, Eyes $20,000 Price — TrendForce (via Google News), Sep 21
• AMD hits $1 trillion market cap for the first time as stock rides 5-day rally — CNBC Tech, Sep 21
• Holy S**t! Ron Baron Put’s Elon’s “99%” Prediction In Perspective — Steven Mark Ryan, Sep 21
• SpaceX IPO Debuts at $135 at a $1.75 Trillion Valuation on June 12 — Where Could SPCX Trad — TradingKey (via Google News), Sep 21
• SpaceX IPO Ignites Space Stocks, Which Stock Is Most Worth Investing In? — TradingKey (via Google News), Sep 21
• Tech stocks today: Oura files IPO paperwork, Nvidia's Huang pushes back on AI doomsday tal — Yahoo Finance (via Google News), Sep 21