AI: Funding Tightens, SpaceX and Anthropic Under Pressure

Key Takeaways

1SpaceX derives more than half of its revenue from AI and Starlink, but shows high Capex and debt

2Anthropic significantly increases its revenue but records massive net losses

3AI funding is contracting, with a drop in debt issuance and more cautious banks

4Open source models are advancing, with Muse, Reflection, and NVIDIA leading the rankings

Why it matters : The financial viability of the AI ecosystem is weakened by rising capital needs, credit contraction, and uncertain economic results.

Credit is tightening for AI infrastructure as stakes soar. SpaceX reports revenue boosted by AI and Starlink but faces record Capex and high debt, while Anthropic posts massive losses. Meanwhile, open source is advancing with new rankings and hardware partnerships.

SpaceX Bets on AI and Starlink with Record Capex

SpaceX states that its Super Intelligence and AI infrastructure segment accounts for 52% of its revenue. The company reports $14.58 billion in revenue for the third quarter of 2026, with a net profit of $4.4 billion. It claims 800,000 weekly paying users for Grok Bot, more than double compared to the previous month. Starlink subscribers increased by 1.6 million during the quarter, reaching 13.6 million, with a year-on-year growth of 69%. At the same time, investments are skyrocketing: Capex reached $19.6 billion in Q3 2026, compared to $6.9 billion in Q3 2025 and $18.37 billion in Q2 2026. Debt stands at $38.8 billion, with 78% owned internally, primarily by Elon Musk. The historical launch business generated only $1.77 billion in the third quarter. The company is spending heavily and carrying significant debt, to the point that it is no longer presented merely as a launch company but is almost entirely dependent on AI and Starlink.

Anthropic Increases Revenue but Deepens Losses

In the first half of 2026, Anthropic generated $16.3 billion in revenue, surpassing its total revenue of $4.6 billion for the year 2025. At the same time, the company reported a net loss of $80.7 billion in the first six months of 2026, after recording a loss of $42 billion in 2025. On the infrastructure side, Anthropic is a client of SpaceX's Neo Cloud division dedicated to AI computing and is among its largest anchored clients, utilizing the full capacity of the Colossus 1 data center in Memphis.

AI Credit Tightens: Falling Debt and Cautious Banks

According to the Financial Times, the issuance of AI-related debt fell to $23 billion in September 2026, down from $113 billion in June. Securing financing is expected to become more difficult by the end of 2026 and into 2027, with a potential AI funding crisis being discussed. In 2026, JPMorgan CEO Jamie Dimon confirmed a tightening of credit amid moratoriums on data centers and increased caution regarding concentration risks. About $18 billion in bank loans for an Oracle campus have become stressed. The NVIDIA-backed provider Firmus has shelved an IPO aiming for a valuation of over $30 billion. Looking ahead, Gartner estimates that over $4 trillion in AI infrastructure investments could be subject to depreciation by the end of 2032 if the current pace continues, and predicts that more than 80% of specialized Neo Clouds are unlikely to remain independent by 2030. The volatility of players like Oracle, CoreWeave, and Softbank is thus drawing more attention.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Open Source: New Rankings and Hardware Bet with Apple

The Beam model from Reflection ranks second among open-weight models in the United States with a score of 30 on the AA scale, behind Meta's Muse Spark Open at 39, and ahead of NVIDIA's Nemotron 3.5 Super VL at 26. These results are based on measurements taken on the AA scale. Muse has partnered with Apple, and according to Muse, Muse Spark Open is designed to leverage the Mac Studio equipped with the M5 Ultra chip, with the entire setup presented as the most powerful solution for running a robust model on a desktop computer. However, it is not established that this level of performance will be sufficient to counter the arrival of Chinese open-weight models.

Fundamentals Under Pressure and Unresolved Macro Debate

The gap is widening between the cost of AI infrastructure and the price paid by companies for its use, while a chasm opens between infrastructure spending and expected returns. The rising demand for computing is weighing on the free cash flows of major tech companies. A persistent gap between AI spending and generated revenue is noted, with no clear horizon for closure. The turning point observed in October 2026 regarding data center financing occurs in this context. Furthermore, the assertion that AI is inflationary is accompanied by the idea that it would push the accessibility crisis to new heights in the United States.