Skip to main content

AI infrastructure economics

How the AI buildout is paid for and what it costs to run. Capital spending, GPU-backed debt, datacenter commitments, depreciation and token prices, read from filings and primary sources.

24 articles and 5 videos. Last updated September 8, 2026.

AI infrastructure economics: articles

  • September 8, 2026

    The Cache Is the Price

    Routing a job to a cheaper model looks like free money. Whether it is depends on how the frontier bills the handover.

  • August 25, 2026

    The Host Pays the Empire

    AI data centers are the foreign bases of the US–China tech rivalry, except that the host now pays the rent.

  • July 29, 2026

    Amazon Priced the Frontier and Declined It

    On 30 June, AWS published two announcements pointing in opposite directions. In July, it closed its AGI Lab. All three were the same decision.

  • July 6, 2026

    The Backstop Has a Name Now - part 2

    How Nvidia finances a cloud tells you what it thinks the cloud is worth. CoreWeave and Nebius got equity. Sharon AI got a revenue-share.

  • July 4, 2026

    The Backstop Has a Name Now - part 1

    Nvidia is handing back tens of billions to shareholders. Now it's offering to finance the customers who can't afford its chips.

  • June 12, 2026

    Cash Flow Lends. Valuation Doesn't.

    Three borrowers, three answers, seventy-two hours. The AI debt boom did not slow down last week. It got priced.

  • June 1, 2026

    The Overbuild Put

    Meta is the only hyperscaler without a cloud business. It just told shareholders it might need one, and that is the most revealing thing it has said about a…

  • May 6, 2026

    The $500 Billion Umbrella

    OpenAI called it the largest infrastructure project in history. Now they call it an umbrella.

  • May 4, 2026

    The Round Trip

    Four hyperscalers reported $700 billion in combined AI infrastructure spending. The number that matters is how much of their reported AI revenue is their own investment coming home.

  • April 7, 2026

    Train, Deploy, Write Down

    The AI industry's most expensive assets depreciate faster than anything in corporate history. The accounting rules haven't noticed.

  • April 3, 2026

    Your Parents Paid

    NVIDIA built the world's most profitable hardware company by treating its consumer GPUs as a recruitment pipeline. Now many recruits are buying Macs.

  • March 25, 2026

    The King's New Datacenters

    On February 11, 2025, at Station F in Paris, Emmanuel Macron announced that France had secured €109 billion in private investment for artificial intelligence. He compared the figure favorably to Am...

  • March 24, 2026

    When the Credit Market Builds a Fire Exit

    On Monday, Bloomberg reported that JPMorgan launched a basket of credit default swaps on five hyperscalers, Alphabet, Amazon, Meta, Microsoft, and Oracle, tradeable in $25 million increments.[1] ...

  • March 17, 2026

    Built the Datacenter, Lost the Brain

    On March 10, 2026, Yann LeCun's new company announced a $1.03 billion funding round at a $3.5 billion pre-money valuation.[1] The company had twelve employees, no product, and no revenue. Its techn...

  • March 11, 2026

    Cloud vs. Clout

    Oracle has produced the most leveraged cloud infrastructure transformation on record: $125 billion in debt, $553 billion in contractual commitments, 57 percent of which are concentrated in a single...

  • March 8, 2026

    Welcome to Hotel Abilene

    You can check in. But can you ever leave? OpenAI's $110 billion round and the lock-in dynamics binding OpenAI, Amazon, Nvidia, Oracle, Microsoft, and Meta into a self-recycling AI infrastructure boom.

  • March 1, 2026

    Chip and Mortar

    On Friday morning, Amazon announced a $50 billion investment in OpenAI, $15 billion upfront, another $35 billion conditional on milestones that reportedly include an IPO or the achievement of AGI ...

  • February 27, 2026

    Compute Equals Commitments

    Published before market open on February 27.

  • February 20, 2026

    Two Markets, One Asset: The GPU Debt Market Is Building the Architecture of Its Own Crisis

    On February 11, 2026, Moody's awarded its first-ever Aaa rating to a data center securitization, an $830 million deal from Compass Datacenters backed by six hyperscale facilities with 100% lease r..

  • February 18, 2026

    What a GPU Debt Crisis Would Look Like

    In late August 2025, Magnetar Capital, a hedge fund based in Evanston, Illinois, managing approximately $26 billion in total assets [1], began taking an unusual position in its largest position. If...

  • February 14, 2026

    Jensen's COMECON: How Nvidia Built an Empire of Captive Clouds

    In January 2026, Nvidia invested an additional $2 billion in CoreWeave, bringing its total stake to $5.3 billion and to 11.5% of the company. The press release used the language of partnership: "de...

  • February 12, 2026

    The $3 trillion bet: who wins when everyone builds the pipes?

    The Big Five tech companies plan to spend roughly $700 billion on AI infrastructure in 2026 alone. Cumulative hyperscaler AI capex from 2024 through 2029 is on track to reach $2.5 to $3 trillion. T...

  • February 10, 2026

    Alphabet's 100-Year Bond: A Bet on Immortality

    Alphabet is planning to issue a 100-year bond. Let that sink in for a moment. A company that didn't exist 28 years ago is asking investors to lend it money until 2126.

  • August 24, 2025

    The AI Crash That Isn't: A Reality Check from the Trenches

    The tech media cycle has found its latest drama: "GPT-5 is a failure," "AI is going to crash," "The bubble is bursting." The swamp is restless again, churning out hot takes and doom predictions wit...

AI infrastructure economics: videos

About the author

Julien Simon is an AI Operating Partner at Fortino Capital. He was VP and Chief Evangelist at Arcee AI, Chief Evangelist at Hugging Face and Global Evangelist for AI and machine learning at AWS, and wrote “Learn Amazon SageMaker” (Packt). He has given 693+ talks in 37 countries and published 502+ technical articles. See his experience and Wikidata entry.

Method: figures come from filings, court documents, vendor documentation and measurements I ran, and each piece dates its numbers.

Go further

More topics

Frequently Asked Questions

Who analyzes AI infrastructure economics and capital expenditure from primary sources?
Julien Simon covers ai infrastructure economics in 24 articles and 5 videos on julien.org. The newest articles are "The Cache Is the Price" (September 8, 2026); "The Host Pays the Empire" (August 25, 2026); "Amazon Priced the Frontier and Declined It" (July 29, 2026). How the AI buildout is paid for and what it costs to run. Capital spending, GPU-backed debt, datacenter commitments, depreciation and token prices, read from filings and primary sources.
Where are Julien Simon's pieces on ai infrastructure economics published?
Articles appear first in The AI Realist newsletter (airealist.ai) and are archived on julien.org. Videos are on the YouTube channel @juliensimonfr, each with a full transcript on julien.org.
Which videos does Julien Simon have on ai infrastructure economics?
5 videos on the YouTube channel @juliensimonfr, each with a transcript on julien.org. The newest are "Run performant and cost-effective GenAI Applications with AWS Graviton and Arcee AI" (November 20, 2024); "Deploying SuperNova-Lite on Inferentia2: the best 8B model for $1 an hour!" (September 19, 2024); "Deep Dive: Optimizing LLM inference" (March 11, 2024).