On a live-streamed podcast on September 28, 2026, ARK Invest’s Cathie Wood put a cost figure and a timeline on what AI could do to drug development — and the arithmetic inside her own sentence is worth reading carefully.
Every figure below is a forecast Cathie Wood spoke on a live-streamed podcast. It is not a study, it is not audited and it is not peer-reviewed, and nothing below takes a view on whether it is right. The numbers were read from YouTube auto-generated captions rather than a corrected transcript, so the transcription may be imperfect.She describes the $2.4 billion baseline as including failures, and she gives no failure or attrition rate anywhere in the passage — none is supplied here. She names no source for that baseline, so no study or institution is named below. All percentages and multiples are this publication’s arithmetic on her figures.
On the Moonshots episode uploaded on September 28, 2026, Cathie Wood — chief executive and chief investment officer of ARK Invest — spoke about AI’s role in drug development. The passage, as rendered in the YouTube auto-generated captions on the live stream (and those captions have not been corrected against a verified transcript), reads: “the AI is going to reduce the number of failures and is going to cut the cost to… discover develop a new drug from $2.4 billion including failures to 600 to 700 million and the time is going to drop from 13 years to eight years or fewer… I do believe healthcare is the most profound application.”
These are remarks made on a live-streamed podcast, not in a paper, a filing, or a peer-reviewed publication. That context is not a footnote — it is the frame for everything that follows. Wood names no source for the $2.4 billion baseline in the passage captured here, and no study, author, or institution is named in this piece either. No failure rate, no phase-by-phase success probability, no date or horizon for when any of this would occur, no named drug, programme, company, or trial, and no definition of what “discover develop” covers appears in the passage — and none is supplied here from any external source.
The figures are taken from auto-generated captions on a live stream rather than a corrected transcript. The raw caption renders the baseline as “$2.4 4 billion” and the target as “600 to700 million.” Both ends of her range are carried throughout this piece. No midpoint is invented.
The key insight: The load-bearing phrase in Wood’s forecast is “including failures.” A per-drug cost that counts every programme that died along the way is dominated by those failures, not by the one that worked. So the cost reduction she describes is, mechanically, a statement about failure rates — and no failure-rate figure appears in her passage.
She gave a range rather than a point, so both ends are shown. The baseline is the figure that carries the failures.
The Structural Read
Notice the order of Wood’s clauses. She says the AI “is going to reduce the number of failures and is going to cut the cost” — failures come first, cost comes second — and the baseline she names is explicitly “including failures.” That sequence is not incidental.
When a per-drug cost figure is constructed to include every programme that failed along the way, the figure is dominated by those failures rather than by the single approved drug at the end. Cutting that aggregate by roughly three and a half times is, mechanically, a statement that a large share of today’s failures stop happening. This piece’s own arithmetic: $2.4 billion to $700 million is a reduction of approximately 70.8 per cent, or roughly 3.4 times. To $600 million it is 75 per cent, or four times. Both ends are carried; no midpoint is chosen.
The timeline arithmetic is different in character. Thirteen years to eight is five years shorter, a reduction of approximately 38.5 per cent. In Wood’s account, cost falls by three to four times while duration falls by under forty per cent. This piece records that asymmetry without resolving it. No mechanism is offered to explain the gap, and no view is taken on whether the combination is plausible.
“The AI is going to reduce the number of failures and is going to cut the cost to… discover develop a new drug from $2.4 billion including failures to 600 to 700 million and the time is going to drop from 13 years to eight years or fewer… I do believe healthcare is the most profound application.”
As rendered in YouTube auto-generated captions. Not a corrected transcript. Raw caption reads “$2.4 4 billion” and “600 to700 million.”
Harness Theory distinguishes between companies that build AI capabilities and companies that harness them to restructure existing cost structures. Wood’s framing maps cleanly onto the harnessing side: the claim is not that AI produces better drugs, but that it eliminates the programmes that would otherwise fail. The evidence required to evaluate that claim is attrition data — before-and-after phase success rates — not cost-per-approved-drug figures alone. That data does not appear in her passage, and is not supplied here.
How to read the cost figure
A per-drug cost built to include failures cannot be reduced by making the surviving drug cheaper to run. It can only fall if fewer programmes fail, or if each failed programme becomes cheaper — most plausibly by dying earlier. Wood’s phrasing covers both channels — “reduce the number of failures and… cut the cost” — which is why the distinction matters. But with no failure-rate figure attached to the forecast, the relative contribution of each channel is unspecified, and this piece supplies no figure for either.
The second thing worth recording is an asymmetry that falls out of the arithmetic. On her own numbers cost falls by three to four times while duration falls by under forty per cent, and those two figures arrived in a single sentence. No explanation for the gap is offered here and no view is taken on whether it is coherent — that would mean engaging with mechanisms the passage does not contain.
The third is about what kind of statement this is. Wood is chief executive and chief investment officer of ARK Invest, an asset manager, and the remarks were made on a live-streamed podcast — which makes them an investment thesis rather than a research result, and the standard of evidence appropriate to the first is not the standard appropriate to the second. Nothing here states or implies what ARK owns, holds, has bought or recommends. No fund, position or price target is named, and nothing here suggests the thesis is self-interested.
Wood’s forecast is a thesis about harnessing AI to restructure a coststructure built around failure rates — not about building AI. The Map of AI traces exactly where in the stack that kind of value capture happens, and which layers have to move for a claim like this to become observable. If you want the structural vocabulary to think about forecasts like this one, that is where to start.
Cathie Wood’s forecast on the Moonshots livestream of September 28, 2026 is a single spoken passage by an asset manager on a podcast — not a study, not audited, not peer-reviewed, and built on a baseline whose source she does not name in the clip. The figures, read from auto-generated captions and carried here at both ends of her range without a midpoint, point to a reduction in drug-development cost of between 70 and 75 per cent and a timeline reduction of roughly 38.5 per cent. The structural observation — that a cost figure “including failures” is driven by attrition, and that the failure-rate data required to underwrite the claim does not appear in the passage — is narrower than a verdict on whether she is right. This piece takes no such verdict. The observation is simply that her own phrasing locates the supporting evidence in a place she does not go.
Source: Cathie Wood, Moonshots livestream, YouTube, September 28 2026. Figures read from auto-generated captions; no corrected transcript has been used. All percentages and multiples are this publication’s own arithmetic on Wood’s stated figures.
Every figure above is a forecast Cathie Wood spoke on a live-streamed podcast. It is not a study, it is not audited and it is not peer-reviewed.The numbers were read from YouTube auto-generated captions rather than a corrected transcript, so the transcription may be imperfect — the raw caption renders the baseline as $2.4 4 billion and the target as 600 to700 million. She describes the $2.4 billion baseline as including failures, and she gives no failure or attrition rate anywhere in the passage. None is supplied above from any other source.She names no source for that baseline, and no study, author, institution or dataset is named above — guessing which published estimate she had in mind would be invention rather than reporting. All percentages and multiples are this publication’s own arithmetic on her figures. She gave a range rather than a point, so both ends of it are carried above and no midpoint is computed. Nothing above states or implies what ARK Invest owns, holds, has bought or recommends; no fund, position or price target is named; and nothing above suggests the thesis is self-interested. The only affiliation fact used is that she is chief executive and chief investment officer of an asset manager, which makes the remarks an investment thesis rather than a research result. Where a cost figure that includes failures is described as making this a claim about the failure rate, that is a reading of her own phrasing and is expressly not a rebuttal. This piece takes no view on whether the forecast is right. Any failure or attrition rate, any date or horizon for when this would occur, any named drug, programme, company or trial, any regulatory position on AI-derived evidence, and any peer-reviewed support are not established and do not appear — a limit of one spoken passage and of this reporting rather than evidence that none exist. Nothing above predicts anything in this publication’s own voice. Nothing here is investment advice.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
