Video Summary

How OpenAI ends and takes Oracle with it | Ed Zitron

The Tech Report

Main takeaways
01

Token‑based billing is becoming unavoidable as AI providers burn subsidies and face unsustainable compute costs.

02

Anthropic's $1 trillion secondary‑market valuation is likely detached from operational reality and liquidity.

03

Microsoft moving GitHub Copilot to token billing signals even well‑capitalized players are cutting subsidies.

04

OpenAI projects huge revenues but also needs hundreds of billions in funding; failure could drag Oracle down.

05

Project Stargate is behind schedule, risks GPU obsolescence, and Oracle depends on its success to meet cash needs.

Key moments
Questions answered

Why does Ed Zitron say token-based billing is unavoidable?

Because AI providers are subsidizing massive compute costs and burning billions; moving to per‑token charges is the only practical way to stop unsustainable losses and reveal true unit economics.

How is Anthropic’s $1 trillion valuation characterized?

As detached from reality: it’s based on thin, low‑liquidity secondary trades rather than public market volumes and likely overstates operational value.

What does Microsoft moving GitHub Copilot to token billing imply?

It implies even highly profitable, well‑capitalized companies feel pressure to cut subsidies and align pricing with compute costs, potentially reducing users' willingness to pay.

How does Project Stargate tie Oracle’s fate to OpenAI?

Oracle has heavily invested in Stargate and expects revenue from supplying OpenAI; delays, limited hardware, or OpenAI failing to meet funding targets could leave Oracle unable to cover its negative cash flow.

What GPU risks could trigger wider industry problems?

Delays and rapid GPU generation cycles make deployed hardware obsolete, canceled orders and legal/supply issues (e.g., Super Micro) can create oversupply of GPUs, collapsing secondary prices and straining debt‑backed data centers.

Token-Based Billing as a Necessity 00:00

"Eventually, everyone will have to do token-based billing. There’s no avoiding it unless they want to just continue annihilating billions of dollars."

  • Companies like OpenAI are under financial pressure to switch to a token-based billing system to avoid massive monetary losses.

  • OpenAI forecasts a potential revenue of $284 billion by 2030, yet they would need to raise $218 billion in funding or debt to sustain operations.

  • The staggering financial figures published by credible sources indicate an unstable financial situation for these AI companies.

The Illusion of Valuation in AI Startups 00:51

"Anthropic has overtaken OpenAI and has hit a $1 trillion valuation, at least on the secondary markets."

  • The valuation of companies like Anthropic is deemed extremely detached from reality due to low liquidity and trading volumes in private secondary markets.

  • Many perceive these inflated valuations as speculative with no substantial proof justifying their worth, unlike traditional stocks in the public market.

  • The recent lack of excitement around OpenAI suggests a shift in focus to Anthropic, but the sustainability of such valuations is questionable.

The Absence of Profitability in AI Operations 02:48

"What would happen if Anthropic charged for the actual running costs of their AI or implemented their underlying business model without subsidization?"

  • There is skepticism regarding the profit margins of AI companies, with estimates suggesting gross margins might actually be far lower than claimed.

  • Operating under a subscription model may not be feasible without transitioning to token-based billing, which some believe could reveal the true financial status of these companies.

  • With no proven profitability and business models heavily reliant on subsidies, there are concerns about their long-term sustainability.

Microsoft's Move to Token-Based Billing 04:36

"Microsoft is moving all GitHub Copilot subscribers to token-based billing."

  • Microsoft's change in billing practice for GitHub Copilot signifies a larger trend within the industry, as they recognize the need to lower operational costs.

  • The decision to shift to token-based billing may hint at financial strain even for a well-capitalized company like Microsoft, which could impact user willingness to continue subscriptions.

  • Analysts are questioning whether customers will accept this change, given that it fundamentally alters the cost structure of services they previously viewed as subsidized.

The Inevitable Testing of the AI Bubble 06:57

"If they stop subsidizing the underlying compute, suddenly we’ll find out whether they actually matter."

  • The transition to token-based billing can serve as a critical test for the viability of the AI bubble, revealing whether users value AI products at their actual cost.

  • The potential backlash from consumers unfamiliar with this billing structure suggests that many may not respond positively to these changes, indicating weak market demand.

  • This shift could expose the overhyped nature of AI technologies that have thrived during periods of heavy subsidization, prompting a reevaluation of their true market value.

Pricing Concerns and Rate Limits 09:14

"OpenAI is just going out there saying, 'We've reset rate limits again.'"

  • There is an increasing trend where companies are not adequately explaining their price hikes, which range from 30% to 100%. This situation is creating unease in the market.

  • OpenAI has communicated undefined resets of rate limits, which seems to encourage users to spend more without much justification for how the pricing structure will work in the future.

  • The transition towards token-based billing appears inevitable unless companies wish to continue incurring massive financial losses.

Nvidia's GPU Claims Under Scrutiny 09:56

"They said their new Ver Rubin GPUs will be reducing the energy usage per token by up to 35 times."

  • Nvidia claims that their new GPUs can reduce energy consumption significantly, asserting a 35-times cheaper rate for token usage.

  • The credibility of such claims is questionable, as they often rely on specific benchmark tests using open-source models, leaving room for skepticism regarding practical applications and actual costs.

  • Transparency regarding operational costs for AI models like those from OpenAI or Anthropic is lacking, making it challenging to validate any claims about cost efficiency.

The Shift to Token-Based Billing 11:35

"The inherent unreliability of large language models means that you can't say with any guarantee how much or how long anything will take."

  • As user bases for AI services decrease, companies might face budget evaluations that question the sustainability of their AI investments.

  • There's growing concern about the possible emergence of a "technofeudalism" where a small segment benefits from AI while the wider workforce does not, though this narrative isn't necessarily true.

  • Many users are not accustomed to a token-based economy, and reports indicate that spending on AI services can quickly become excessive, prompting organizations to reassess their AI budgets.

Financial Implications of AI Utilization 13:57

"There were some businesses spending 10% of their headcount on AI tokens, and that could increase to 100%."

  • Some businesses are reportedly spending a significant portion of their budget on AI, which may lead to unsustainable financial practices.

  • The expectation was that AI would replace labor costs, yet companies are finding themselves investing more in tokens than they would in employees, contradicting the initial rationale.

  • Firms are beginning to recognize the financial implications of AI usage and will likely take a step back to evaluate whether the costs justify the productivity gains claimed by AI solutions.

Market Sentiment and Future of AI Companies 14:49

"Once one of the major labs goes full token-based, that's how you know it's over."

  • There is rising unease in investing in AI companies as capacity issues, deteriorating product quality, and lack of transparency regarding performance metrics arise.

  • The narrative surrounding AI's value proposition is becoming less convincing, leading to skepticism about future investments in the sector.

  • As organizations contemplate moving to token-based billing, the narrative of AI's remarkable potential may diminish, resulting in a moment of reckoning for companies at the forefront of AI development.

The Inconsistencies of AI Services 18:39

"It’s the worst possible bubble we could ask for. It loses the most money and the products are dull."

  • The speaker critiques the performance of AI companies like Anthropic, describing their services as inconsistent with uptime issues, even suggesting that Anthropic is overvalued despite its operational failures.

  • They express frustration over the notion that users must pay more for worse service in these AI platforms, arguing that the overall experience is disappointing and frustrating for end-users.

  • The commentary spans broader concerns about the AI bubble, highlighting a perceived disconnect between inflated market valuations and actual product effectiveness.

Problems with Project Stargate 22:10

"There's barely any hardware in there. OpenAI needs to take residency of this so Oracle can pay its project management fees."

  • Discussion shifts to Project Stargate, revealing significant delays in the construction of data centers and emphasizing that only a fraction of the planned infrastructure is operational.

  • The financial implications for Oracle are dire; they are relying heavily on the successful completion of Stargate to maintain their fiscal health, which is entangled with OpenAI's performance and commitment.

  • The projections for revenue from this project are optimistic, yet the reality on the ground shows a stark contrast to expectations, raising concerns about Oracle's future viability without these revenue streams.

Financial Dependencies and Risks 26:07

"Oracle cannot pay its bills. Oracle had negative cash flow of nearly $25 billion last quarter."

  • The financial state of Oracle is precarious, with a significant negative cash flow and heavy reliance on OpenAI to fulfill their economic obligations.

  • If OpenAI fails to meet its projected revenue targets, Oracle could face severe financial repercussions, indicating a risky business model based on contingent income.

  • The commentary also highlights the potential fallout for Larry Ellison, whose personal financial stakes in Oracle exacerbate the dire warnings about the company’s future should their contracts with OpenAI fail.

Issues with Data Center Timelines and GPU Obsolescence 28:25

"Everything that you're hearing about data centers is already obsolete."

  • The ongoing construction challenges for data centers have raised significant doubts about their ability to keep pace with evolving GPU technology. As new GPU generations are released, existing hardware is quickly becoming outdated.

  • For instance, the GPUs that will ultimately be employed in the Stargate project are expected to be two years old by the time they are installed, rendering them nearly obsolete upon deployment.

  • Additionally, the introduction of newer GPUs, such as those from the Vera Rubin family, adds to the complexity and potential for misalignment between infrastructure development timelines and GPU releases.

  • The energy requirements and thermal cooling needs for newer GPU models have further compounded these issues, making it impractical to simply replace old hardware with new without substantial physical modifications to data center infrastructure.

The Impact of Super Micro's Challenges on Oracle 31:46

"Oracle canceled about 1.1 to 1.4 billion dollars worth of chips from Super Micro."

  • Oracle’s significant cancellation of chip orders with Super Micro signals deeper issues within the industry. This might suggest a lack of trust or readiness in the supply chain, as Oracle may be reassessing its hardware needs amid obsolescence concerns.

  • The legal troubles faced by Super Micro, including the arrest of co-founder Wally VR for selling GPUs to Chinese entities, have likely contributed to Oracle's decision to cancel orders. This situation raises questions about Super Micro's reliability in the long term.

  • Because Super Micro has a substantial inventory of outdated GPUs that cannot be sold, this creates a precarious financial situation for the company. With an annual revenue of around ten billion dollars, a billion dollars worth of unsold inventory represents a serious burden.

The Potential GPU Market Collapse 33:51

"GPUs will simply kind of sit in it because Nvidia's not going to buy them back."

  • As data centers struggle to meet current demands and adapt to new technologies, the prospect of an oversupply of GPUs in the market threatens to cause a significant collapse in GPU value.

  • There's speculation that if these GPUs were to flood the secondary market, it could result in drastic losses across the industry, particularly for companies that leveraged substantial debt tied to GPU assets.

  • The situation is exacerbated by the fact that current rental rates for GPUs are failingly low, revealing that many centers might not even cover their operational costs.

  • If financial pressures continue to mount and data center operators begin to collapse, a wave of refinancing efforts could ensue as companies rush to stabilize their operations while they are still afloat.

Oracle's Financial Challenges and Future Prospects 37:41

"At some point, someone's going to take too big a loss and they're going to run out of money."

  • Oracle is currently in a precarious financial position, as they attempt to sell their compute resources aggressively, even if it results in losses. This strategy raises concerns about sustainability, as continued losses could lead to financial failure.

  • The viability of Oracle's future is in question, with bankruptcy being a potential outcome unless they can recoup costs, particularly from their partnership with OpenAI. The likelihood of Oracle sinking alongside OpenAI is a genuine concern.

  • Strategically, Oracle might have the option to exit certain projects, such as the Stargate data centers, which are housed under separate entities. They could explore legal avenues, potentially even refusing to pay their debts.

  • Legal battles seem to be a familiar territory for Oracle, which is noted for its robust legal team, often being described as a "law firm with software attached."

  • Currently, Oracle's growth is largely driven by demand for AIG GPUs. Without this segment, the rest of their business struggles to grow significantly.

  • The company faces a daunting challenge as OpenAI reportedly needs a staggering $852 billion over the next four years, an amount that emphasizes the gravity of their financial landscape.

  • Oracle’s options appear limited. If they cannot secure the necessary funding or revenue, they might have to dismantle parts of their business and turn to other tech giants, like Microsoft, for survival.

  • The potential for Oracle to not build additional data centers exists, which could mitigate some debt risks, but they have already deeply committed financially, complicating their exit strategies.

  • Overall, the outlook for Oracle, especially in relation to OpenAI, is bleak, with speculation that they may have to divest from other major investments, such as CBS and their TikTok shares, to stay afloat.