“Took almost 3 years, many billions of dollars (over a half-trillion fieldwide).”
Our reading:
GPT-5 took almost three years and many billions of dollars, with over half a trillion dollars spent fieldwide.
Assessment: Unresolved
What happened
Was it true for the period described?
Outcome: Unresolved
We checked, and the evidence so far doesn’t settle it.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The release dates allow an elapsed-time reading without requiring a development start. ChatGPT to GPT-5 spans about two years eight months; GPT-4 to GPT-5 spans about two years five months. Marcus did not select the baseline. Reported OpenAI compute spending establishes a multibillion-dollar scale, but does not allocate that spending to GPT-5. The reviewed sources do not establish a comparable fieldwide total above US$500 billion by launch. With the financial scope and accounting basis unspecified, the combined claim cannot fairly be scored supported or contradicted.
Caveats
- Unknown accounting basis prevents a supported or contradicted financial score under the stated rules.
- The spending context relies on a secondary synthesis of reports and estimates, not a public audited GPT-5 cost disclosure.
- Failure to establish the fieldwide total is not evidence that spending was below the threshold.
Evidence (4 items)
Context only: OpenAI,
a model called ChatGPT
Why it matters: Establishes the initial ChatGPT release as one plausible baseline. The announcement does not say GPT-5 development began then.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- About events on
- Published
Context only: OpenAI,
We are releasing GPT‑4’s text input capability
Why it matters: Establishes GPT-4's release as another plausible baseline. Marcus did not explicitly select it.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- About events on
- Published
Context only: OpenAI,
We are introducing GPT‑5
Why it matters: The endpoint gives 981 elapsed days from ChatGPT's release and 877 from GPT-4's release. These are about two years eight months and two years five months, respectively.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Math
- 2025-08-07 minus 2022-11-30 = 981 days; 2025-08-07 minus 2023-03-14 = 877 days. (inputs: e1: ChatGPT release, 2022-11-30; e2: GPT-4 release, 2023-03-14; e3: GPT-5 release, 2025-08-07; Day counts are exact; year-and-month descriptions omit remaining days.)
- About events on
- Published
Context only: Epoch AI,
billions of dollars on compute
Why it matters: Documents reported multibillion-dollar company-wide compute spending before launch. Its scope does not identify how much belongs to GPT-5 or establish the fieldwide total.
- Publisher
- Epoch AI
- Kind of source
- Secondary summary
- Measures
- OpenAI, Annual cloud compute spending, USD dollars, estimate, 2024-01-01 to 2024-12-31
- About events on
- Published
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Conditions
- almost 3 years (their words)
- over a half-trillion fieldwide (their words)
Where we had to interpret
- Read as an approximate elapsed-time claim ending at GPT-5's launch, accompanied by claims about spending incurred by then.
- Marcus did not identify the starting milestone. Consider both ChatGPT's initial release and GPT-4's release without requiring a documented development start.
- Do not impose an exact numerical tolerance on 'almost 3 years'.
- Read 'many billions' as multiple billions spent on the effort leading to GPT-5, without limiting that effort to its final training run or assuming all OpenAI expenditure belongs to it.
- Read 'over a half-trillion fieldwide' as more than US$500 billion of actual cumulative AI spending. The field's boundaries and accumulation start are unspecified.
- Leave the target-period start unspecified because the post supplies no starting milestone or spending baseline.
- No accounting basis or financial source is identified. Record the basis as unknown.
Right if: A defensible elapsed-time baseline consistent with the context, together with comparable spending records, establishes approximately three years, multiple billions attributable to the effort leading to GPT-5, and more than US$500 billion actually spent across the stated field by launch. Financial support requires resolving the unspecified accounting basis and scope.
Wrong if: Comparable records establish that a stated component is false under its intended baseline and scope. A narrower spending subtotal, a different release baseline, or missing disclosures alone cannot contradict the claim.
We wrote this test after the outcome was known.
What would settle it
A clarified elapsed-time baseline and financial scope, plus records allocating actual spending to the effort leading to GPT-5 and a consistently measured cumulative industry total through August 7, 2025.
We found 10 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, GPT-5 hot take, paragraph 3.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2025-08-07.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, GPT-5 hot take
Took almost 3 years, many billions of dollars (over a half-trillion fieldwide).
- Where
- paragraph 3
- Published
- 2025-08-07
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Unresolved. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
A reviewer said our test wasn’t fair to what was actually said.