Michael Saylor has offered a striking example of how artificial intelligence may be entering a field usually associated with investment bankers, securities lawyers and corporate finance teams. In an August 2026 interview with Steven Bartlett on The Diary of a CEO, Saylor described using ChatGPT while working through the development of STRK, a preferred security ultimately issued by Strategy. The security is real, its financial terms are documented, and investors supplied real capital. The more dramatic interpretation of the story requires considerably more care. [1]
The public record does not establish that ChatGPT independently designed STRK, that human professionals could not have produced the structure without it, or that STRK itself raised $15 billion. Strategy's SEC filings show that STRK's initial offering generated approximately $563.4 million in estimated net proceeds. They also reveal a much larger capital architecture involving several preferred securities, common equity and other financing mechanisms. [2]
The distinction matters beyond Saylor or Strategy. It provides an unusually clear case study in how AI may influence the development of sophisticated financial products while still operating inside the human legal, regulatory and capital-markets system required to make those products real.
What Did Michael Saylor Say About ChatGPT and STRK?
Saylor's account is important because he is a direct source for his own process. In the Diary of a CEO interview published August 6, he discussed using ChatGPT while reasoning through financial structures that eventually became part of Strategy's capital strategy. [1]
That establishes an important but limited fact: Saylor says he used ChatGPT during the development process.
A founder is ordinarily in a strong position to describe the tools he personally used, the questions he asked and the ideas he explored. The evidentiary problem begins when testimony about process is converted into proof of causation. Public filings can confirm what STRK ultimately became, how it was registered, how much capital particular offerings produced and what rights investors received. They cannot independently reconstruct which participant originated each design idea inside Strategy or determine whether ChatGPT was indispensable to the final result.
The strongest version of the story is therefore not that ChatGPT autonomously invented a financial instrument. It is that the executive chairman of a public company says he used a general-purpose AI system while developing a real financial product that subsequently passed through the conventional machinery of regulated capital markets.
That is consequential without requiring the stronger claim.
What Is STRK?
STRK is Strategy's 8.00% Series A Perpetual Strike Preferred Stock. It is a genuine registered security that trades on Nasdaq and is convertible into Strategy Class A common stock under its governing terms.
MicroStrategy, as the company was then named, priced the initial STRK offering on January 30, 2025. The company offered 7.3 million shares at $80 per share and estimated approximately $563.4 million in net proceeds after underwriting discounts, commissions and estimated offering expenses. The security carried a $100 liquidation preference and an 8% annual cumulative dividend. [2]
Those details are important because they move the story from AI experimentation into regulated finance. STRK was not a hypothetical structure generated inside a chatbot conversation. It became a security with investors, contractual rights, dividend obligations, conversion provisions, underwriting and SEC disclosure.
At the same time, the SEC record places a hard boundary around claims about how much money STRK initially raised.
Did STRK Raise $15 Billion?
Not in its initial offering.
Strategy's SEC filing puts estimated net proceeds from the original STRK offering at approximately $563.4 million. [2]
The confusion becomes easier to understand when Strategy's broader financing system is examined. In March 2025, the company established an at-the-market program authorizing as much as $21 billion of additional STRK issuance. An authorization of that size is not the same thing as receiving $21 billion from investors.
Later SEC disclosures make the distinction explicit. During the third quarter of 2025, Strategy reported approximately $152.8 million in net proceeds from STRK shares sold through the ATM program. Between October 1 and October 26, it reported another approximately $23.8 million. At that point, approximately $20.4 billion of STRK issuance capacity remained available. [3]
In other words, a large ATM authorization describes how much a company may potentially issue under the program. It should not be reported as money already raised.
That distinction makes a simple "$15 billion with ChatGPT" narrative difficult to sustain if the number is being attributed to STRK alone. The independently documented initial STRK offering was much smaller, while Strategy's overall financing architecture extends across multiple securities and programs.
Strategy Built More Than One Preferred Security
STRK also makes more sense when viewed as one component of a broader financial architecture rather than as an isolated product.
Strategy has issued several separate preferred-stock classes with different economic structures, including STRF, STRC, STRK, STRD and the euro-denominated STRE. SEC records identify these as distinct securities rather than variations of one STRK offering. The STRE filing records a separate initial public offering that produced approximately 608.8 million euros in net proceeds, reported at roughly $703.9 million using the exchange rate stated in the filing. [4][5]
Strategy has also used common-stock issuance and other financing mechanisms as part of its capital strategy. This larger structure is important when interpreting any broad statement about billions of dollars being raised.
The accurate financial story is not that one ChatGPT-assisted security suddenly produced the entire amount. Strategy constructed a multi-instrument capital system over time, with STRK occupying one important position within it.
Where Does AI Actually Fit Into the Process?
This is where Saylor's account becomes more interesting than the headline version.
A sophisticated financial product begins as an idea, but an idea alone cannot become a publicly traded security. The final instrument has to survive legal review, financial analysis, corporate approval, regulatory disclosure, underwriting and ultimately investor judgment.
That means any ChatGPT involvement described by Saylor necessarily existed within a much larger institutional process.
AI could potentially help a founder explore structures, compare possibilities, question assumptions or iterate through ideas much faster than traditional research alone. That would be a meaningful change in executive capability if it proves reproducible. It does not mean the AI becomes the securities lawyer, investment bank, accountant, board, regulator or investor.
STRK ultimately carried formal dividend terms, liquidation rights, conversion provisions and SEC disclosures. It was underwritten and sold into an actual market. Those facts demonstrate the continued role of the human and institutional system surrounding whatever AI-assisted ideation occurred beforehand.
The more defensible interpretation is therefore that AI may have expanded the founder's ability to explore financial structures before conventional institutions reviewed and executed them.
Did ChatGPT Design STRK?
Public evidence cannot answer that question with enough precision to say yes.
Saylor can credibly report that he used ChatGPT and describe the importance he personally assigns to that process. SEC filings can independently confirm the resulting security.
What those filings cannot reveal is the internal intellectual history of every provision. They do not identify which STRK terms originated with Saylor, which emerged through conversations with AI, which came from bankers, which were revised by lawyers or which evolved through negotiations among multiple parties.
The evidence therefore supports saying that Saylor used ChatGPT during STRK's development if his first-person account is being reported accurately. It does not independently establish that ChatGPT designed the final security.
That difference between testimony and outcome evidence is important whenever founders discuss AI-assisted accomplishments.
Was ChatGPT Indispensable?
That claim is even harder to establish.
To show that ChatGPT was indispensable, evidence would have to demonstrate more than participation. It would have to establish that the resulting structure could not reasonably have been produced without ChatGPT or an equivalent system.
No public evidence reviewed for this report establishes that counterfactual.
This does not mean Saylor is wrong about the importance AI had to his own thinking. A tool can transform how an individual works without being provably necessary to the final outcome.
The narrower claim is also arguably more interesting. If AI allows a founder to participate much more deeply in technical financial design before professional specialists take over, it may change who can meaningfully contribute to the earliest stages of financial engineering.
That possibility can be studied without claiming that lawyers or bankers have become unnecessary.
What Strategy's Bitcoin Holdings Show About the Scale of the Company
The financial system surrounding these securities has become enormous.
A Strategy SEC filing dated July 6, 2026, reported approximately 843,775 bitcoin held as of July 5. Compared with Bitcoin's fixed maximum protocol supply of 21 million coins, that represents approximately 4.02%. [6]
The distinction between maximum protocol supply and currently circulating Bitcoin is important. Saying Strategy held roughly 4% of Bitcoin's maximum possible supply is more precise than saying it owned 4% of all Bitcoin, because not all 21 million coins have been mined and some already-mined coins may be permanently inaccessible.
Why Strategy Selling Bitcoin Matters
The same filing also revealed an important evolution in Strategy's capital system. Between June 29 and June 30, the company sold 1,363 BTC for approximately $80.8 million. Between July 1 and July 5, it sold another 2,225 BTC for approximately $135.2 million. Strategy stated that proceeds from those sales were used to fund distributions on its preferred stock and replenish the portion of its U.S. dollar reserve used for that purpose. [6]
That creates a more complex picture than the familiar narrative of Strategy continually issuing securities simply to acquire more Bitcoin. Capital is now moving in both directions.
Strategy's financing architecture was widely understood through a relatively simple loop: raise capital, acquire Bitcoin and use the company's access to public capital markets to expand its holdings.
The July 2026 filing demonstrates that the system can also operate in reverse. Bitcoin can be monetized to support obligations created by the securities surrounding it.
Preferred securities create distributions that must be funded. Debt creates interest obligations. A dollar reserve creates another source of liquidity, and Bitcoin itself can become a monetizable asset when cash is required.
The same filing reported a U.S. dollar reserve of $2.55 billion and described the reserve as supporting preferred-stock dividends and interest on outstanding indebtedness. [6]
The Larger AI Story Is Financial Engineering
Saylor's story points toward a larger question that extends well beyond Strategy.
Generative AI has already demonstrated that it can help users search large bodies of information, compare documents, explore alternatives, draft language and reason interactively through complicated problems. Financial-product design involves many of those same activities, but it also operates inside a heavily regulated environment where mistakes can create enormous consequences.
That creates a potentially important division of labor.
AI may increasingly become part of the exploratory layer of finance, where founders, bankers, analysts and lawyers examine structures and iterate through possibilities. Human professionals and institutions remain responsible for validating those ideas, determining whether they comply with applicable law, translating them into binding documents and deciding whether they should actually reach investors.
STRK does not prove that this transition has occurred across finance. It provides a real example of a prominent corporate executive publicly describing AI as part of his own financial-design process.
That makes the broader phenomenon worth watching.
The Problem With Founder Causal Compression
There is also a broader reporting lesson in how stories like this spread.
When a founder says that a particular technology helped produce a major result, several separate causal stages can quickly collapse into one sentence. AI was used. A financial product was developed. Investors bought securities. The company raised capital. The capital supported a larger corporate strategy.
Those events may all be true without the first event being the sole cause of everything that follows.
The problem is not unique to Saylor. Technology stories frequently compress a chain of contributors into the most interesting component because the compressed version makes a stronger headline.
For financial reporting, that compression can become materially misleading.
A more accurate account of STRK recognizes both sides of the story. Saylor's testimony is legitimate evidence that he used ChatGPT as part of his process. Strategy's filings are stronger evidence for what the resulting financial instrument actually was and how much money particular offerings generated. The two forms of evidence answer different questions.
What This Case Actually Establishes
The evidence establishes something narrower than the viral version but potentially more consequential over the long term.
Michael Saylor says he used ChatGPT while developing a real financial product. Strategy subsequently issued STRK through the regulated U.S. capital markets. The initial offering produced approximately $563.4 million in estimated net proceeds, and the company later established a much larger ATM authorization under which additional STRK shares could be sold over time. [2][3]
What the evidence does not establish is that ChatGPT independently created STRK, that human financial professionals could not have produced it without AI, or that STRK alone raised $15 billion.
That boundary should not make the story less interesting. It makes the real question clearer.
If executives can use increasingly capable AI systems to explore financial structures before lawyers, bankers and regulators formalize them, AI may begin changing financial engineering long before it replaces any of the institutions responsible for making finance work. STRK may be an early example of that shift. Whether it becomes a widespread practice is now a question worth investigating.