@michaeljburry here is what he’s showing 🚨
OpenAI, xAI, Mistral, etc., don’t need 6–10 gigawatts of GPU to serve consumers. They’re not building user platforms. They’re building infinite training sinks where they can burn investor dollars in real-time on Nvidia clusters, while claiming “AGI progress.”
•You’re not paying for finished goods you’re paying for ongoing compute consumption, and that consumption is the product.
•Every dollar raised goes to burn GPUs.
•Every GPU burned gets turned into a higher valuation.
•Valuation justifies more raises.
•More raises justify more compute.
This is circular value creation burn → raise → burn → raise. It mimics product momentum, but it’s 100% synthetic.
——
And Nvidia = The New Central Bank of Silicon Valley
Nvidia is no longer just a hardware company it’s acting like a shadow central bank, injecting liquidity into the system by offering compute in exchange for optionality, influence, or future stake.
•Nvidia “invests” $100B into OpenAI? That’s not equity altruism it’s forward-guaranteed CapEx demand.
•Nvidia funds or supplies CoreWeave, Figure, Mistral why? Because Nvidia isn’t just selling chips, it’s selling an entire monetary substrate to which the AI ecosystem is tethered.
GPU allocation is now a monetary instrument. It’s how influence flows. It’s the new capital issuance.
—-
And CapEx Is the Collateral l Not the Output
OpenAI’s $500B valuation? That number is not backed by product revenue. It’s backed by:
•Future GPU burn.
•Option deals with AMD and Nvidia.
•Exclusive compute agreements with Microsoft and Oracle.
•Perceived AI dominance if CapEx continues indefinitely.
It’s not what OpenAI builds it’s what OpenAI commits to spend that drives value.
This flips capitalism on its head:
•Normally: You build value → reinvest capital → scale.
•Now: You raise money → burn CapEx → get value.
This model only works as long as someone believes the future revenue is “inevitable.” The second revenue misses expectations? The entire thing is exposed as CapEx leverage cosplay.
—-
The Ultimate Grift: Renting “Time to AGI” as a Business Model
The illusion of “approaching AGI” is being used as a derivative product. Every startup is now selling some variant of:
“We’ll be the first to reach AGI, and every GPU we consume brings us closer. So fund us to burn more compute faster than the next guy.”
This is the equivalent of Wall Street selling volatility not real returns, just exposure. And the thing they’re selling is the future probability curve of artificial general intelligence, not a product that makes money today.
The longer they can rent belief in AGI’s approach, the more liquidity they can extract without delivering anything of substance.
And the beautiful part for them?
Nobody knows what AGI looks like.
There’s no finish line.
Which means: The burn never has to end.
—-
That public markets will one day absorb these valuations through IPOs or acquisitions.
But:
•Cloud growth is flatlining.
•Retail AI usage is stalling (chatbot fatigue, hallucination risk, enterprise pushback).
•Regulators are waking up (EU AI Act, antitrust in US/UK).
•Electricity/grid costs for GPU farms are exploding.
•And AI monetization is turning out to be way harder than “just plug in GPT.”
So if the IPO window doesn’t open in the next 24 months, this becomes a terminal feedback loop a $trillion dollar treadmill of cash incineration with no exit.
And that’s how it collapses:
•No exit = down rounds.
•Down rounds = CapEx slows.
•CapEx slows = Nvidia growth halts.
•Nvidia halts = the collateral pool dies.
•The entire liquidity mechanism unravels.
@michaeljburry Here is what he’s saying 🚨 Michael Burry was warning us in no uncertain terms that we were being conned.
If we had widespread testing from the start testing the healthy, the mildly sick, and the asymptomatic we would have seen the real fatality rate was tiny, closer to seasonal flu levels for most of the population. He’s saying the fear was based on bad data. Not because the virus lied, but because the testing didn’t show the full picture.
——
His “<0.2% fatality rate” claim was based on a premise almost no one was acknowledging in March 2020:
If you test everyone, you uncover all the asymptomatic and mild cases the “invisible” denominator.
Once you include all those invisible infections in your fatality rate calculation, the scary headlines vanish. The virus becomes far more survivable, especially for the non-elderly.
In other words:
•COVID wasn’t killing 3–5% of people who got it.
•It was likely killing well under 0.2%, but only if you counted everyone who had it not just the sick ones showing up at hospitals.
That difference between case fatality rate (CFR) and infection fatality rate (IFR) is everything.
Burry saw through it early. Most policymakers didn’t. Or worse they did, but ignored it.
—-
He’s saying: if the fatality rate was this low, we didn’t need to crush 99.8% of people’s lives to “save” the 0.2%. The cure was worse than the disease and not by a little, but by orders of magnitude.
The people running the show weren’t trying to minimize total harm. They were trying to minimize blame. That’s why they nuked the economy instead of focusing protection on nursing homes the actual kill zone.
In short: never let a crisis go to waste. The lockdowns weren’t just overreactions they were also opportunities for control, distraction, and economic restructuring that benefitted the already-powerful.
——
PCR Testing Was Weaponized to Inflate “Cases,” Not Clarify Risk
This one’s huge. PCR tests weren’t just diagnostic they became the emotional barometer of the pandemic.
But here’s what most people never grasped:
•PCR tests can detect tiny fragments of viral RNA, even weeks after infection, when the person is no longer contagious or symptomatic.
•The “cycle threshold” (Ct) how many cycles of amplification are run was often set absurdly high (35–45 cycles), which massively increased false positives or clinically irrelevant positives.
•Fauci himself quietly admitted (in offhand interviews) that anything above 33 cycles was essentially junk data for transmission relevance.
Now think about this:
Governments locked down entire cities based on “case spikes” driven by hyper-sensitive PCR tests, not by meaningful increases in actual illness or hospitalizations.
Translation: the “second wave” and much of the panic was statistical theater, not epidemiological reality.
This kept the fear alive and justified extending emergency powers indefinitely.
——
By early-mid 2020, internal data (later leaked from German, UK, and U.S. agencies) showed:
•IFR for people under 40 was well under 0.01%.
•IFR for children and teens was close to zero.
•80+% of deaths were among people over 70, often with comorbidities, and especially those in institutionalized care.
And yet…
•Schools were shut down.
•Outdoor playgrounds were taped off.
•Parks were closed.
•Working-age people were told to stay home.
•“Non-essential” businesses (aka small, independent ones) were obliterated.
None of that matched the age-targeted risk profile of the virus.
Why?
Because a targeted response is too smart to justify sweeping control.
Blanket panic enabled blanket policy which meant maximum authority with minimum nuance.
—-