The third quarter of 2026 represents a foundational paradigm shift for Space Exploration Technologies Corp. (SpaceX). In its first fully reported quarter following its historic $85.7 billion June 2026 initial public offering (IPO), the company's financial disclosures reveal an enterprise that has aggressively transitioned from a pure-play aerospace manufacturer and satellite broadband provider into a highly capitalized, vertically integrated superpower in digital infrastructure and artificial intelligence (AI).

SpaceX reported consolidated revenues of $14.58 billion for the quarter ended September 30, 2026, representing an extraordinary 161.4% year-over-year (YoY) expansion and significantly outperforming consensus estimates of $12.87 billion. The company generated GAAP net income of $4.40 billion—a dramatic swing from a net loss of $1.05 billion in Q3 2025—and delivered Adjusted EBITDA of $8.81 billion, marking a 306% YoY increase. However, the defining revelation of the Q3 print is the structural inversion of the company's revenue base. The Super Intelligence (SI) segment, fortified by the February 2026 acquisition of xAI and the recent $60 billion all-stock integration of Cursor, generated $7.56 billion in the quarter, officially eclipsing the Connectivity (Starlink) segment as the primary engine of corporate growth.

Beneath the headline profitability lies an unprecedented capital expenditure (Capex) cycle. SpaceX deployed $19.59 billion in Capex during Q3 alone, with $16.44 billion (83.9%) directed exclusively toward AI compute infrastructure. This capital intensity illustrates a profound strategic evolution: the legacy aerospace division is increasingly serving as an internal logistics mechanism to deploy orbital telecommunications and compute networks, while terrestrial AI infrastructure consumes the vast majority of the company's liquidity.

Simultaneously, SpaceX is executing aggressive horizontal and vertical integrations that threaten to disrupt adjacent global industries. The $17 billion acquisition of EchoStar's AWS-4 and H-block spectrum signals SpaceX's intent to operate as a primary, facilities-based Mobile Network Operator (MNO), bypassing legacy terrestrial telecom partnerships. Furthermore, the establishment of special board committees to explore a potential merger with Tesla, Inc. sets the stage for a theoretical $3.5 trillion corporate consolidation designed to unify AI compute, autonomous robotics, energy storage, and aerospace under a single corporate architecture.

The Q3 2026 financial print highlights the dual nature of SpaceX's current economic model: hyper-scale top-line growth coupled with aggressive infrastructure spending, all funded by an exceptionally fortified balance sheet.

Consolidated revenue for the three months ended September 30, 2026, reached $14.58 billion, up from $5.58 billion in the prior-year period. For the nine months ended September 30, 2026, revenue totaled $27.08 billion, a 97.5% YoY increase. The cost of revenue scaled at a slower pace than the top line, increasing 102.5% YoY to $5.48 billion in Q3. This dynamic demonstrates significant operating leverage as the company monetized high-margin cloud services and software subscriptions against fixed infrastructure costs.

Operating income turned firmly positive, reaching $2.79 billion compared to an operating loss of $273 million in Q3 2025. However, the discrepancy between the $2.79 billion operating profit and the headline $4.40 billion net income requires rigorous scrutiny. The bottom line was heavily flattered by below-the-line, non-operating items. SpaceX recorded $873 million in interest income, primarily earned on the massive cash pile accumulated from the IPO proceeds. Additionally, the company recognized $1.03 billion in other income, which included a $561 million unrealized gain on equity investments in unconsolidated affiliates—notably stemming from a September 2026 financing round for The Boring Company (TBC). A $390 million provisional tax benefit, triggered by the release of a valuation allowance related to deferred tax liabilities recognized during the Cursor and Mesh Optical acquisitions, further widened the gap between core operating performance and net income. Consequently, approximately 37% of Q3 net income was derived from non-operational accounting mechanics rather than recurring business activities.

Source: SpaceX Q3 2026 Form 10-Q

Research and development (R&D) expenses increased 100.6% YoY to $4.60 billion in Q3, heavily burdened by $1.02 billion in share-based compensation expenses related to replacement equity awards granted during recent acquisitions, alongside escalating infrastructure costs for Super Intelligence data centers and ongoing Starship development. Selling, general, and administrative (SG&A) expenses rose 196.9% to $1.71 billion, similarly impacted by acquisition-related share-based compensation and a substantial increase in international marketing for the Starlink broadband service.

While the income statement reflects a transition to profitability, the cash flow statement reveals a corporation engaged in an unparalleled capital deployment cycle. For the nine months ended September 30, 2026, SpaceX generated $13.10 billion in operating cash flow. However, capital expenditures over the same period reached a staggering $48.07 billion, resulting in a negative free cash flow profile of approximately $34.97 billion.

This profound capital deficit contextualizes the strategic necessity of the June 2026 IPO. The public offering was not merely a liquidity event for early stakeholders; it was a mandatory recapitalization required to fund the company's multi-front infrastructure roadmap. The $85.67 billion in net IPO proceeds, combined with $11.99 billion in net debt issuance, successfully bridged this free cash flow gap while massively fortifying the balance sheet.

As of September 30, 2026, SpaceX boasts exceptional liquidity. The company holds $51.29 billion in cash and cash equivalents, augmented by $38.56 billion in marketable securities, yielding total available liquidity of nearly $89.86 billion. Total debt and finance leases stand at $38.84 billion. Notably, the company's debt profile includes the $25 billion "SpaceX Notes" issued in June 2026—a five-tranche senior unsecured offering with a weighted-average maturity of 11.7 years and a weighted-average coupon of 5.855%. The proceeds from these notes were utilized to retire a highly restrictive $20 billion bridge loan, which had previously been used to extinguish term loans and senior secured notes assumed during the xAI acquisition.

Furthermore, the balance sheet features $12.98 billion in "other financings," predominantly comprising failed sale-leaseback transactions for SI infrastructure hardware orchestrated with related parties, including Valor Equity Partners. Despite this debt load, the sheer magnitude of SpaceX's cash reserves allows management to front-load capital commitments for NVIDIA GB300 clusters, orbital constellations, and terrestrial semiconductor fabrication plants without immediate reliance on operating cash flow.

The Super Intelligence (SI) segment has rapidly eclipsed SpaceX's legacy aerospace and satellite operations to become the primary driver of corporate valuation and capital allocation. Originally formed via the consolidation of xAI and X Corp, and further expanded by the $60 billion acquisition of Cursor and the $249 million acquisition of Mesh Optical, the SI segment reported $7.56 billion in Q3 2026 revenue.

SI segment revenue surged 829.9% YoY from $813 million in Q3 2025. This exponential growth was driven primarily by SI Solutions and Infrastructure, which generated $7.02 billion in the quarter. SpaceX has effectively birthed the "neocloud," monetizing its massive GPU clusters by renting compute capacity to third parties—including a notable $6.3 billion multi-year compute-leasing contract with Reflection AI and a $920 million per month arrangement with Google—at premium spot-market rates.

The August 2026 acquisition of Cursor (Anysphere, Inc.) proved immediately accretive to the top line. Post-acquisition, Cursor contributed $608 million in revenue during the final six weeks of the quarter. The mechanics of the Cursor deal highlight the immense purchasing power of SpaceX's newly minted public equity. The $42.47 billion total purchase consideration was executed almost entirely via an exchange ratio of 9.7142 shares of SpaceX Class A common stock per Cursor share, valuing Cursor at an implied equity value of $60 billion based on a seven-day Volume-Weighted Average Price (VWAP) of $128.65. The acquisition resulted in the recognition of $5.30 billion in intangible assets (predominantly existing technology and customer relationships) and $36.65 billion in goodwill. Even excluding the Cursor contribution, SI infrastructure revenue expanded an impressive 192% sequentially.

Advertising revenue from the X platform, though entirely overshadowed by compute leasing, grew 19% YoY to $538 million. This indicates a stabilization of the platform's core monetization engine following the introduction of a revamped advertising technology stack and higher demand from financial services and technology advertisers. Subscriptions for Grok Bot and enterprise AI agents are also witnessing accelerating adoption, with paid weekly users for Grok Bot surpassing 800,000.

Source: SpaceX Q3 2026 Form 10-Q

The financial toll of maintaining frontier AI model leadership and scaling a global neocloud is staggering. The SI segment absorbed $16.44 billion in Capex during Q3 2026, accounting for 83.9% of the corporation's total quarterly spending. Chief Financial Officer Bret Johnsen projects total company Q4 2026 Capex to range between $35 billion and $40 billion, driven primarily by the ongoing acceleration of data center buildouts. For context, the combined 2026 Capex guidance for the traditional hyperscaler oligopoly (Microsoft, Amazon, Alphabet, and Meta) is estimated at roughly $725 billion; SpaceX's current spending trajectory firmly establishes it as a peer in this ultra-capital-intensive arena.

SpaceX exited Q3 with 1.7 gigawatts (GW) of nameplate compute capacity. CEO Elon Musk outlined an extraordinarily aggressive deployment curve: targeting 2.5 to 3 GW by the end of 2026, and aiming to bring an incremental 10 to 20 GW of power and cooling online in 2027. The Colossus I and II facilities in Memphis, Tennessee, and the Macrohard site in Southaven, Mississippi, serve as the epicenters of this buildout, housing hundreds of thousands of liquid-cooled NVIDIA GB300, GB200, and H100 GPUs.

The economics of these data centers currently justify the immense expenditures. Management noted that current compute deployments are generating payback periods of less than one year, with pricing tracking slightly above the high end of the $30 to $50 per watt range forecasted in prior quarters. The insatiable demand from hyperscalers and AI startups creates a dynamic where SpaceX can dictate terms on short-term spot contracts, capitalizing on the increasing value of "useful intelligence per watt" as base models continue to improve.

A critical third-order insight from the Q3 earnings call involves how SpaceX is systematically overcoming the physical constraints of AI scaling. The limiting factor for gigawatt-scale data centers is no longer just securing silicon allocation, but rather the deployment of power generation, cooling, and electrical equipment.

To bypass a notorious 12-to-18-month supply chain lag for critical gas turbine components, SpaceX has re-tasked its advanced metallurgical casting teams—originally developed for the Raptor rocket engine program—to manufacture blades and vanes for terrestrial gas turbines at a new foundry in Bastrop, Texas. Aerospace manufacturing demands materials that withstand extreme thermal and pressure environments, making the transition to industrial gas turbine components highly synergistic. By insourcing the most complex and bottlenecked components of power generation, SpaceX is fundamentally altering the deployment speed of digital infrastructure, positioning itself to bring gigawatts of power online significantly faster than competing hyperscalers.

The Connectivity segment, driven by the Starlink low-earth orbit (LEO) broadband constellation, delivered a highly profitable and strategically vital quarter. Revenue reached $5.25 billion (up 69.5% YoY and 22% sequentially), generating a robust operating income of $2.37 billion. Operating margins for the segment expanded to 45.1%, up from 38.4% in the prior year. Crucially, Connectivity recorded Segment Adjusted EBITDA of $3.50 billion. After accounting for $1.54 billion in segment Capex, Connectivity cleared approximately $1.96 billion in free cash flow, making it the sole segment within SpaceX that currently funds its own growth while throwing off excess cash to support the broader enterprise.

Starlink ended Q3 with 13.6 million active broadband consumer subscribers across 174 markets, adding 1.6 million consumers in the quarter alone. Consumer Average Revenue Per User (ARPU) landed at $69 per month, reflecting a 5% sequential increase attributed to favorable mix dynamics, though it remains lower on a YoY basis ($83 in Q3 2025) as the company expands aggressively into developing markets (such as Vietnam and Uganda) with localized pricing tiers.

Enterprise and government connectivity revenue surged 91% YoY, highlighting a highly successful pivot toward higher-margin, sticky B2B applications. The aviation sector represents a particularly lucrative growth vector; Starlink is now contracted for approximately 12,000 aircraft (roughly 25% of the global commercial fleet), with major international carriers including Lufthansa, Swiss, Copa Airlines, United Airlines, and Korean Air activating services during the quarter. Management projects that enterprise and government revenue will ultimately exceed consumer broadband revenue in the near future.

Source: SpaceX Q3 2026 Form 10-Q

The successful deployment of 26 Starlink V3 satellites during Starship Flight 14 marks a generational technological leap in orbital telecommunications. The specifications of the V3 architecture alter the fundamental physics and routing geometry of the constellation.

Each V3 satellite is a petabit-class network node weighing approximately 1,500 to 2,000 kg, designed exclusively to fit within the massive payload bay of the Starship vehicle. The technical capabilities represent a massive upgrade over the prior generation:

This optical fabric creates a highly redundant, ultra-low-latency mesh network in Low Earth Orbit (LEO). By routing data via 1550 nm near-infrared laser pulses between satellites at the speed of light in a vacuum (which is roughly 30% faster than light traveling through terrestrial fiber optic glass), Starlink bypasses the need for localized ground station relays. This enables secure, gigabit-speed connectivity over oceans, polar regions, and hostile landmasses. CEO Elon Musk envisions scaling this architecture to a 100,000-satellite constellation within three years, anticipating a 1,000x increase in global bandwidth demand driven by autonomous AI agents that consume exponentially more data than human users.

In a masterstroke of financial engineering and spectrum acquisition, SpaceX executed a definitive agreement to acquire the entirety of EchoStar's prized AWS-4 and H-block mid-band spectrum. The transaction, valued at approximately $19.6 billion, was uniquely structured to address EchoStar's acute liquidity crisis while securing vital radio frequencies for SpaceX. SpaceX will fund up to $2 billion in cash interest payments on EchoStar debt through November 2027, alongside an $11.1 billion equity issuance (261.8 million Class A shares at a fixed value of $42.40 per share) upon the final closing.

This acquisition fundamentally rewrites the competitive landscape of the U.S. telecommunications sector. Historically, satellite-to-cellular (Direct-to-Cell/D2C) models relied on wholesale partnerships with terrestrial carriers (e.g., AST SpaceMobile partnering with AT&T) and utilizing the carrier's existing spectrum allocations. By owning the AWS-4 and H-block mid-band spectrum outright, SpaceX is no longer bound by these partnerships. When combined with the recently approved acquisition of 15,000 Mobile V2 satellites and a low-band 800 MHz spectrum portfolio from Grain Management, SpaceX possesses the complete radio frequency stack required to operate as an independent, facilities-based Mobile Network Operator (MNO).

This strategic pivot poses a severe existential threat to both legacy terrestrial carriers (Verizon, T-Mobile) and nascent D2C competitors. Market reaction was swift and punitive, with AST SpaceMobile shares dropping nearly 14% on the announcement, as investors recognized SpaceX's capacity to bypass carrier roaming agreements and offer direct, native 5G connectivity from space directly to unmodified smartphones.

The foundational Space segment generated $1.77 billion in Q3 revenue, an 84% sequential and 6.1% YoY increase. Despite executing fewer commercial customer launches than the prior-year period, the segment benefited from a highly favorable payload mix and the execution of high-value government contracts. The segment recorded an operating loss of $7 million, heavily skewed by a 44.3% YoY increase in R&D expenses ($1.23 billion) dedicated almost entirely to the accelerated development and iterative testing of the Starship program.

SpaceX conducted 35 orbital launches in Q3 2026, comprising 21 internal Starlink deployments on Falcon 9, one Falcon Heavy (NASA's Roman Space Telescope), 12 external customer missions, and two Starship test flights. The company maintains near-monopolistic control over orbital logistics, accounting for approximately 90% of all global mass-to-orbit, with China representing roughly 6% and the rest of the global launch market combining for a mere 4%. Total mass to orbit for the quarter reached 476 metric tons, down slightly from 525 metric tons in Q3 2025, though 358 metric tons of this volume was dedicated to internal Starlink payloads.

Starship Flight 14 was a pivotal success, marking the first revenue-generating (via internal segment transfer) deployment by placing operational Starlink V3 satellites into orbit. Management anticipates that Flight 15 will attempt the unprecedented capture of both the Super Heavy booster and the Starship upper stage using the launch tower's mechanical arms (the "chopsticks"), achieving full and rapid reusability. Should this succeed, the marginal cost of placing payloads into orbit will collapse, enabling the economic viability of deploying heavy compute payloads, massive space structures, and interplanetary missions.

The external launch manifest is expanding rapidly, with SpaceX securing three additional NASA Commercial Crew missions through 2028, a lunar surface payload contract for Starship, and 18 new National Security Space Launch (NSSL) Phase 3 missions. The U.S. Space Force awards under NSSL Phase 3 Lane 1 and 2—including missions for the Space Development Agency's Tranche 2 Tracking Layer and the NRO's Starshield constellation—lock in billions of dollars in high-margin government revenue through the end of the decade.

A compelling commercial development within the Space segment is the introduction of the "Starfall" microgravity capsule. Designed as an uncrewed, mass-producible reentry vehicle, Starfall utilizes its launch vehicle for orbital insertion and relies on nitrogen cold-gas thrusters and parachute recovery for Pacific Ocean splashdowns, offering short-duration orbital manufacturing environments.

SpaceX has signed milestone contracts with Redwire's SpaceMD and European startup Space Cargo Unlimited to fly dedicated Starfall missions. Space Cargo Unlimited intends to launch a one-tonne space factory utilizing the Starfall platform by 2028, opening orbital access for pharmaceutical crystallization, biotechnology, and advanced materials engineering. Redwire's SpaceMD plans to fly 32 "PIL-BOX" systems on a 2028 Starfall mission to scale in-space pharmaceutical development. By bypassing the logistical bottlenecks, high costs, and vibrational interference of the aging International Space Station (ISS), Starfall effectively commoditizes microgravity, creating a scalable, high-margin "microgravity-as-a-service" market.

Source: SpaceX Q3 2026 Form 10-Q

SpaceX’s most audacious infrastructure bet extends beyond orbital constellations and AI data centers to the foundational layer of computing hardware: semiconductor fabrication. The company's "Terafab" initiative, situated on a footprint exceeding 100 million square feet in Grimes County, Texas (near the Gibbons Creek Reservoir), aims to internalize the production of custom silicon required for SI model training, Tesla's Full Self-Driving (FSD) architecture, and Optimus humanoid robots.

Structured formally as a joint venture, Terafab commands a $16.8 billion capital commitment for Phase 1, with county tax filings projecting up to $119 billion in total investment across multiple phases. The physical scale of the project is unprecedented; upon completion, it is designed to be the largest building on Earth by floor area, eclipsing Russia's AvtoVAZ assembly plant, the Pentagon, and Giga Texas. The facility's operational success relies heavily on deep strategic partnerships with Intel (via the Intel 14A node framework) and ASML.

During the earnings call, Musk noted that ASML is expediting Extreme Ultraviolet (EUV) lithography equipment deliveries to support the project. SpaceX targets manufacturing a functional static random-access memory (SRAM) test chip at its Austin research fab by mid-2027, with volume production at the main Grimes County facility aspirated for Q3 or Q4 2028.

The strategic implication of Terafab cannot be overstated. By attempting to vertically integrate advanced chip fabrication, SpaceX aims to break the geopolitical and supply-chain stranglehold currently maintained by TSMC and NVIDIA. Should Terafab achieve commercial viability at scale, SpaceX transitions from being the world's largest consumer of AI silicon to an autonomous producer, cementing a hardware moat that no software-centric hyperscaler (e.g., Google, Microsoft, Meta) can easily replicate.

During the Q3 earnings call, Elon Musk confirmed that both SpaceX and Tesla have formed special committees of independent directors to formally evaluate a corporate merger. A combination of the two entities would forge a unified conglomerate with an estimated valuation of $3.5 trillion.

The strategic rationale for the merger is undeniably robust. Both companies share immense overlapping dependencies in AI hardware architectures, custom silicon design, and autonomous robotics. Tesla's expertise in manufacturing scale, battery technology (specifically the Megapacks used extensively to stabilize power at SpaceX's data centers), and edge inference computing aligns perfectly with SpaceX's Terafab semiconductor ambitions, orbital infrastructure, and frontier SI model training via xAI. Unifying the entities eliminates redundant R&D expenditures, prevents internal competition for capital and engineering talent, and consolidates Musk's cognitive and temporal bandwidth.

However, the mechanical execution of the merger faces formidable corporate governance and structural hurdles. At SpaceX, Musk wields an iron grip, controlling over 82% of the voting power through dual-class super-voting (Class B) shares, making approval on the SpaceX side a mere formality. Conversely, Musk owns approximately 20% of Tesla's outstanding equity. Musk explicitly specified on the earnings call that a successful merger requires the affirmative vote of 50% of all outstanding Tesla shares, not merely a majority of the votes cast.

This high absolute threshold transforms shareholder apathy into active dissent; if a retail shareholder abstains or a broker does not vote, it effectively counts as a "no" against the merger. The precedent is concerning for proponents of the deal: a recent Tesla proposal to scrap supermajority voting rules drew 73.6% participation, yet only 39.4% of outstanding shares voted "yes," resulting in failure.

Furthermore, the exchange ratio mechanics present a severe dilution of voting power for Tesla shareholders. Because SpaceX is a dual-class entity, a stock-for-stock acquisition using standard one-vote Class A shares would effectively transfer Tesla’s assets into a corporate vehicle completely dominated by Musk’s 10-vote Class B shares. In a theoretical at-market deal, Tesla's outside holders—who currently control roughly 80% of Tesla's vote—would see their voting influence in the combined company plummet to approximately 11%, while Musk’s voting control over the combined entity would surge to approximately 73%. This stark transfer of governance control will undoubtedly be scrutinized heavily by institutional investors and proxy advisory firms, making the special committee negotiations highly contentious.

The aggressive, multi-front scaling of SpaceX's operations has attracted intense legal and regulatory scrutiny across multiple jurisdictions, introducing material risks to the corporate balance sheet and operational timelines.

Wall Street's reaction to the Q3 2026 print reflects a rapid consensus re-rating of SpaceX. The market is increasingly valuing the firm not merely as a space logistics and satellite broadband provider, but as a foundational, vertically integrated AI utility. Following the earnings release, the stock surged over 9% in after-hours trading to $161.56, pushing the company's market capitalization well past the $2.1 trillion mark.

Institutional analysts have aggressively adjusted their models to account for the velocity of the SI segment's revenue ramp. JPMorgan issued a highly bullish note, doubling its price target to $500 and projecting a theoretical $6.6 trillion market cap. The bank forecasts SpaceX revenue to reach $254 billion by 2027, driven almost entirely by the targeted 10-13 GW of computing capacity coming online. Goldman Sachs raised its price target to $230 from $220, reflecting confidence in the multi-channel funnel of orbital and terrestrial revenue.

However, the current valuation relies heavily on extreme future execution. Musk's guidance of achieving $100 billion in Annualized Recurring Revenue (ARR) based on December 2026 figures implies a monthly run rate of over $8.33 billion exiting the year. Furthermore, his projection of $500 billion to $1 trillion in ARR by the end of 2027 borders on the unprecedented. To justify the current $2.16 trillion valuation at a mature 5x Price-to-Sales multiple, SpaceX must generate over $430 billion in sustainable revenue by the end of the decade. The execution risks associated with scaling Terafab, deploying 100,000 V3 satellites, securing 20 GW of localized power generation, and navigating the treacherous corporate governance waters of the Tesla merger represent the critical variables in these financial models. The divergence in analyst sentiment is stark, with Morningstar maintaining a conservative fair value estimate of $63, highlighting the immense speculative premium currently priced into the stock.

SpaceX's Q3 2026 financial results crystallize a masterclass in industrial evolution and capital deployment. By leveraging the cash flow generation of its Starlink broadband monopoly and the vast capital access afforded by public equity markets, the company has funded a hostile, vertically integrated takeover of the artificial intelligence infrastructure layer. The Super Intelligence segment is no longer an ancillary venture; it is the financial engine of the enterprise, capturing 84% of capital expenditures and driving the overwhelming majority of revenue growth.

The strategic moats currently being excavated are generational. The EchoStar spectrum acquisition fundamentally transitions Starlink into a global, facilities-based mobile network operator, bypassing terrestrial telcos and neutralizing nascent D2C competitors. The Starlink V3 optical mesh network effectively creates an orbital data center architecture that routes traffic independent of vulnerable terrestrial chokepoints. Meanwhile, the audacious Terafab project and the insourcing of gas turbine component manufacturing demonstrate a refusal to be constrained by legacy semiconductor and energy supply chains.

Moving into 2027, the primary catalyst for SpaceX will be the resolution of the Tesla merger negotiations. Should the merger succeed—overcoming the steep 50% outstanding voting threshold required of Tesla shareholders—Elon Musk will have forged a unified, multi-trillion-dollar conglomerate controlling the vertical stack of the 21st-century economy: clean energy generation, orbital logistics, global telecommunications, autonomous robotics, custom silicon fabrication, and frontier artificial intelligence.

The immediate financial imperative for SpaceX is executing the transition of its massive $63 billion revenue backlog and gigawatt-scale data center capacity into sustainable, long-term free cash flow. The company must achieve this before the massive liquidity buffer provided by the 2026 IPO is depleted by the relentless, compounding capital demands of the global AI arms race.