We checked the viral "three things just broke" argument against the government's own numbers

An argument is circulating that three things broke at once this month, and that together they show a reset being quietly switched on. The emergency oil reserve is being drained. The Treasury has started buying its own debt. And the gains from all of it are being funnelled to companies ordinary investors cannot buy.

It is a coherent thesis, it is well told, and the man telling it is selling a webinar and an app subscription at the end. None of that makes it wrong. So we checked twelve of its factual claims against the source data, most of which is published free by the US government.

Some of it is exactly right. Four claims are wrong by a large margin. And the most interesting thing is which way the errors run.

What holds up

The oil reserve numbers are correct to the barrel. The Strategic Petroleum Reserve stood at 284.552 million barrels in the week to 18 September, the lowest since October 1982, after exactly 26 consecutive weekly declines and a 405,000 barrel fall that week. We checked every one of those against the EIA series and they are right.

What the argument leaves out is why, and we covered that separately: it is an authorized exchange, not a quiet depletion, and exchanged barrels carry a legal obligation to come back. His strongest single point is also the least discussed one, that Venezuelan crude physically cannot meet the reserve's specification.

The Treasury did raise its buyback size to $4 billion. That is real, announced on 19 August. The description of what a buyback is, however, is wrong at every step, which we set out in a separate piece. Treasury is the buyer, not the seller. The Federal Reserve is not a counterparty. The cash comes from bill issuance. No new money is created.

The five largest companies really are about 30% of the S&P 500. On the SPDR holdings file for 23 September: Nvidia 8.22%, Apple 7.41%, Microsoft 5.60%, Alphabet 5.38% across both share classes, Amazon 3.68%. That is 30.3%, against 27.2% two years ago. At the dot-com peak the top five were around 18%.

The money supply claim is right. M2 went from $15.49 trillion in February 2020 to $21.79 trillion in March 2022. That is a 40.6% increase in 25 months.

And the trades are real. The president's June disclosure, filed with the Office of Government Ethics, runs to more than 1,000 trades worth between $78.1 million and $263.1 million. The purchases include Berkshire Hathaway, Visa, Mastercard, Cintas and Home Depot. Meta was sold. The pattern described, dull cash businesses bought and crowded technology sold, is visible in the filing.

What does not

"Central banks are buying gold at the fastest pace since 1997." Wrong year and wrong direction. Central banks were net sellers through the 1990s and 2000s and did not become net buyers until 2010. The record year was 2022 at 1,082 tonnes, which the World Gold Council called the highest annual demand on record back to 1950. Buying in 2025 was 863 tonnes, still far above the 473 tonne average of 2010 to 2021, but slowing rather than accelerating.

"In the 1970s gold went up eight times." It went up considerably more. Gold was fixed at $35 an ounce until the gold window closed in August 1971 and reached $512 by the end of 1979, which is 14.6 times. To the London fix of $850 on 21 January 1980 it is 24.3 times.

"A 1971 dollar is worth 7 cents now, according to the US government." The government's own calculator says otherwise. The BLS calculator puts $1.00 from January 1971 at $8.42 in August 2026 money, which makes the 1971 dollar worth about 12 cents, not 7.

"Inflation officially hit 11%." US consumer price inflation peaked at 9.1% in June 2022, the largest twelve month increase since 1981. It never reached 11% at any point. Producer prices did approach 11.7% in March 2022, so there is a real number nearby, but it is not consumer inflation and the argument is about what things cost households.

"SpaceX, Anthropic and OpenAI are worth more than every US company that went public in the last 45 years combined." This one fails twice over.

It is wrong by roughly a factor of ten. Nvidia alone, which listed in 1999, is worth about $5.41 trillion. Apple, Alphabet, Microsoft, Amazon, Meta, Broadcom and Tesla take eight post-1980 listings past $25 trillion between them. The three private companies are worth around $3.8 trillion on confirmed marks.

And SpaceX is not private. It listed on Nasdaq on 12 June 2026, raising $85.7 billion including the greenshoe at a $1.77 trillion valuation, the largest IPO ever held. A claim whose premise is "three companies you and I cannot buy" includes one that has been publicly traded for three months.

There is a real statistic underneath it, and it is striking on its own: the three do exceed the $4.07 trillion combined first-day market value of all 3,365 US technology IPOs between 1980 and 2025, from Jay Ritter's dataset at the University of Florida. Drop the words "technology" and "first-day value" and it becomes false.

The errors do not run one way, and that is the point

The natural assumption about a video selling an inflation webinar is that every number will be inflated. That is not what we found.

He understates the two claims that would have helped him most.

Gold in the 1970s went up 14.6 times, not eight. If your argument is that currency debasement rewards hard assets, the real number is nearly twice as good as the one you used.

And the concentration claim, the one genuinely worth worrying about, is materially worse than he says. He puts household exposure at "about a quarter" of net worth. The Federal Reserve's Z.1 accounts for the second quarter of 2026 put corporate equities plus mutual fund holdings at 35.3% of household net worth, against 26.2% at the dot-com peak and 21.7% before the financial crisis. Counting only directly held shares gets you 26.6%, which is presumably where the quarter comes from, but the number everyone else quotes is more than a third. It is an all-time high on every measure in a series going back to 1945.

Meanwhile the claims that make the best hooks, 7 cents and three companies beating 45 years of listings, are the ones that overshoot.

That is the signature of an argument assembled from secondary summaries rather than from the underlying data. Not dishonesty. Repetition, of numbers that had already drifted before they were picked up.

What an owner or investor should actually take from it

• The concentration risk is the real finding, and it is worse than presented. If your pension sits in an index fund, more than 30% of it is five companies, and your household sector has never had more of its net worth in equities. That is checkable, it is a record, and it does not require any theory about a reset.
• The oil reserve is thinner but not unexplained. A 40% full tank with 130 million barrels out on loan is a smaller shock absorber. It is not a secret.
• Treasury buybacks are a maturity swap, not money printing. The legitimate worry is rollover risk from shortening the average maturity of the debt, which is a real argument and a different one.
• Check the number before you act on it. Every figure in this piece came from the EIA, the Treasury, the Federal Reserve, the BLS, the World Gold Council or a university dataset. All free, all public, and four of them contradicted a claim being repeated to hundreds of thousands of people.

Sources checked 24 September 2026: Federal Reserve Z.1 table B.101 for 2026 Q2; SPDR S&P 500 holdings file for 23 September 2026; World Gold Council Gold Demand Trends; the BLS CPI inflation calculator and CPI release of 13 July 2022; FRED series M2SL; Jay Ritter's IPO statistics at the University of Florida; EIA series WCSSTUS1; and US Treasury press release sb0607. Nothing here is investment advice, and we hold no position in anything mentioned.