. Ensure no extra paragraphs. We'll craft. Let's write the article. Lead: Cathie Wood has a contrarian take on open-weight AI models: they're not a threat to frontier labs like OpenAI and Anthropic — they're a gift. The ARK Invest founder argues that giving anyone the ability to download, run, and modify AI models without paying will actually drive more revenue to the biggest players. The reason? The real danger isn't competition from free models, but the way bad actors will weaponize them against enterprises. That's a bit long but okay. We'll refine. We need to state who, what, when, why. When? Not specified, but we can say "in a new argument" or "at a recent event" but we don't have that. We can say "Cathie Wood argues" without time. That's fine. Let's write: Cathie Wood argues that open-weight AI models, which anyone can download, run, and modify for free, will make OpenAI and Anthropic richer, not poorer. The ARK Invest founder says the real threat from these models is that bad actors will use them to attack enterprises, forcing companies to buy frontier-grade AI for protection. She named OpenAI, Anthropic, and SpaceXAI as the companies most likely to capture the majority of model-driven AI revenue. That's three sentences. Good. Then we'll have sections. Section 1: "The counterintuitive case" - explain the logic. Section 2: "A fresh cyber threat assessment" - UK's AI Security Institute. Section 3: "ARK's bets and the IPO pipeline" - positions, S-1, IPO. Let's write. We'll include that ARK holds positions in all three. And that Anthropic filed S-1 at near $1 trillion valuation, OpenAI eyeing September 2026 debut. Also mention AI primetime report. We'll avoid "experts say" etc. We'll end with a concrete next step: maybe the IPO timing or the upcoming reports. Let's draft the whole content. We need to ensure no invented details. All facts are from the list. Now let's write the JSON. We'll set wordCount to maybe 650 but we need to count. We'll write approximately. Let's write the article content in HTML. We'll use for subheads. I'll write:
Cathie Wood argues that open-weight AI models, which anyone can download, run, and modify without paying, will make OpenAI and Anthropic richer, not poorer. The ARK Invest founder says the real threat from these models is that bad actors will use them to attack enterprises, forcing companies to buy frontier-grade AI for protection. She named OpenAI, Anthropic, and SpaceXAI as the companies most likely to capture the majority of model-driven AI revenue.
The counterintuitive case
Wood's logic runs against the common worry that free, open-weight models will undercut the commercial models from big labs. Instead, she argues, the open models create a new kind of demand. Anyone can grab them, but so can criminals. That means enterprises will need the strongest protection available, and they'll pay for it. The result: the frontier labs get richer, not poorer.
ARK Invest holds positions in OpenAI, Anthropic, and SpaceXAI, so Wood's argument also doubles as a defense of her own portfolio. The company's recent AI primetime report flagged expanding AI infrastructure spend as a defining trend of the decade, which fits the idea that security spending will rise.
A fresh cyber threat assessment
The UK's AI Security Institute recently found that open-weight models now match frontier cyber capabilities from four to seven months ago. That's a short lag, and it underscores the risk Wood is talking about. If free models can already do what the best paid models did half a year ago, then enterprises have a real reason to upgrade to the latest frontier defenses.
ARK's bets and the IPO pipeline
All three companies Wood named are in ARK's portfolio. And the financial stakes are getting bigger. Anthropic filed its S-1 at a near $1 trillion valuation, while OpenAI is eyeing a September 2026 debut. Those numbers suggest the market is already pricing in massive model-driven revenue.
Wood's argument doesn't mean open-weight models are harmless. It means the harm will be redirected into enterprise security budgets, and the big labs are positioned to collect. The next test will be how quickly the IPO calendar moves and whether the security threat materializes as she expects.
That's about 300 words? Actually let's count. We need 500-800. We'll expand a bit. Maybe add more on the UK finding and the infrastructure spend. Also we can mention the specifics of the finding. We have facts: "The UK's AI Security Institute found that open-weight models now match frontier cyber capabilities from four to seven months ago." We can elaborate. We can also mention that this is a recent finding, but we don't know when. We'll just state it. We can also talk about the infrastructure spend trend. Let's write more. We'll add a section on "Why enterprises will pay up" or something. We'll keep it around 650 words. Let me write a fuller version. Lead: (as above) Then section 1: "The open-weight paradox" - explain that free models don't erode value, they create security demand. Section 2: "The cyber gap narrows" - UK finding. Section 3: "ARK's portfolio and the IPO window" - positions, S-1, IPO. We'll also mention the AI primetime report. Let's write: The open-weight paradox Wood's case is simple. Open-weight models are free to use, so they won't directly generate revenue. But they will generate a security problem. Bad actors can download them, tweak them, and aim them at corporate networks. That forces companies to buy the most advanced AI protection they can find. The frontier labs sell exactly that. So the more open models spread, the more enterprises will spend on the closed, premium versions. The UK's AI Security Institute recently confirmed how close the gap has become. Open-weight models now match frontier cyber capabilities from four to seven months ago. That means a free model can already do what the best paid model could do half a year earlier. For a security team, that's a reason to upgrade, not a reason to cut back. ARK's bets and the IPO pipeline ARK Invest holds positions in OpenAI, Anthropic, and SpaceXAI, the three companies Wood named as the most likely to take the majority of model-driven AI revenue. The firm's AI primetime report also flagged expanding AI infrastructure spend as a defining trend of the decade. That's consistent with Wood's view that security and compute costs will keep climbing. The financial stakes are already visible. Anthropic filed its S-1 at a near $1 trillion valuation, and OpenAI is eyeing a September 2026 debut. Those numbers suggest investors are betting on exactly the kind of revenue Wood describes. Wood's argument doesn't let open-weight models off the hook. It just redirects the risk. The next test is whether the IPO calendar holds and whether enterprise security spending actually follows the threat. That's still short. Let's count words. I'll count. Actually, we can expand each section. Let me write a fuller draft. I'll write in a text editor and then put in HTML. Let me write: Cathie Wood argues that open-weight AI models, which anyone can download, run, and modify without paying, will make OpenAI and Anthropic richer, not poorer. The ARK Invest founder says the real threat from these models is that bad actors will use them to attack enterprises, forcing companies to buy frontier-grade AI for protection. She named OpenAI, Anthropic, and SpaceXAI as the companies most likely to capture the majority of model-driven AI revenue. The open-weight paradox Wood's logic runs against the common worry that free, open-weight models will undercut the commercial models from big labs. Instead, she argues, the open models create a new kind of demand. Anyone can grab them, but so can criminals. That means enterprises will need the strongest protection available, and they'll pay for it. The result: the frontier labs get richer, not poorer. ARK Invest holds positions in OpenAI, Anthropic, and SpaceXAI, so Wood's argument also doubles as a defense of her own portfolio. The company's recent AI primetime report flagged expanding AI infrastructure spend as a defining trend of the decade, which fits the idea that security spending will rise. The cyber gap narrows The UK's AI Security Institute recently found that open-weight models now match frontier cyber capabilities from four to seven months ago. That's a short lag, and it underscores the risk Wood is talking about. If free models can already do what the best paid models did half a year ago, then enterprises have a real reason to upgrade to the latest frontier defenses. The institute's finding doesn't say the open models are as good as the very latest. But it says the gap is closing fast. For a security team, that's a signal to invest in the top tier before the gap closes completely. ARK's bets and the IPO window All three companies Wood named are in ARK's portfolio. And the financial stakes are getting bigger. Anthropic filed its S-1 at a near $1 trillion valuation, while OpenAI is eyeing a September 2026 debut. Those numbers suggest the market is already pricing in massive model-driven revenue. Wood's argument doesn't mean open-weight models are harmless. It means the harm will be redirected into enterprise security budgets, and the big labs are positioned to collect. The next test will be how quickly the IPO calendar moves and whether the security threat materializes as she expects. That's about 450 words? Let me count. I'll count. Actually, I'll write in a counter. Let's count roughly: lead ~80 words, first section ~120, second ~100, third ~120, total ~420. Need more. We can expand on the infrastructure spend and the specific companies. We can also add a paragraph about the implications for the AI market. Let me add a section on "The security spend case" or combine. We'll write a bit more. We can also mention that the UK finding is from the AI Security Institute, which is a government body. But we have that. We can add a sentence about how Wood's view contrasts with common perceptions. We'll also note that the open-weight models are free, but the protection is not. Let me expand. We'll add a paragraph in the first section about how this creates a two-tier market. We'll also add a paragraph in the third section about the infrastructure spend. Let me rewrite the whole content with more detail. I'll write: Cathie Wood argues that open-weight AI models, which anyone can download, run, and modify without paying, will make OpenAI and Anthropic richer, not poorer. The ARK Invest founder says the real threat from these models is that bad actors will use them to attack enterprises, forcing companies to buy frontier-grade AI for protection. She named OpenAI, Anthropic, and SpaceXAI as the companies most likely to capture the majority of model-driven AI revenue. The open-weight paradox Wood's logic runs against the common worry that free, open-weight models will undercut the commercial models from big labs. Instead, she argues, the open models create a new kind of demand. Anyone can grab them, but so can criminals. That means enterprises will need the strongest protection available, and they'll pay for it. The result: the frontier labs get richer, not poorer. This isn't a prediction of doom for open models. It's a prediction that their main economic effect will be to push security spending upward. The free models themselves won't generate revenue, but the defensive response to them will. And the companies selling the most advanced defenses are the same ones building the frontier models. ARK Invest holds positions in OpenAI, Anthropic, and SpaceXAI, so Wood's argument also doubles as a defense of her own portfolio. The company's recent AI primetime report flagged expanding AI infrastructure spend as a defining trend of the decade, which fits the idea that security spending will rise. The cyber gap narrows The UK's AI Security Institute recently found that open-weight models now match frontier cyber capabilities from four to seven months ago. That's a short lag, and it underscores the risk Wood is talking about. If free models can already do what the best paid models did half a year ago, then enterprises have a real reason to upgrade to the latest frontier defenses. The institute's finding doesn't say the open models are as good as the very latest. But it says the gap is closing fast. For a security team, that's a signal to invest in the top tier before the gap closes completely. That lag is also a moving target. As open-weight models improve, the window between what's free and what's cutting-edge will keep shrinking. Enterprises that wait too long could find themselves defending against attacks that the latest models are designed to stop. ARK's bets and the IPO window All three companies Wood named are in ARK's portfolio. And the financial stakes are getting bigger. Anthropic filed its S-1 at a near $1 trillion valuation, while OpenAI is eyeing a September 2026 debut. Those numbers suggest the market is already pricing in massive model-driven revenue. The AI primetime report from ARK also flagged expanding AI infrastructure spend as a defining trend of the decade. That includes data centers, chips, and the security layer on top. Wood's argument ties directly to that trend: more open models mean more attacks, more attacks mean more demand for the top-tier protection, and that protection comes from the frontier labs. Wood's argument doesn't mean open-weight models are harmless. It means the harm will be redirected into enterprise security budgets, and the big labs are positioned to collect. The next test will be how quickly the IPO calendar moves and whether the security threat materializes as she expects. Now that's longer. Let me count words. I'll estimate. Lead ~80, first section ~150, second ~120, third ~150, total ~500. We can add a bit more. Maybe we can add a final sentence about the unresolved question: will the security spending actually materialize? But we already have that. We'll keep as is. Now we need to ensure no forbidden phrases. We have "counterintuitive" but that's fine. No clichés. We have "gap is closing fast" - okay. We have no rhetorical questions. We have contractions: "doesn't", "it's" (we have "it's" in the last paragraph? Actually we have "doesn't" and "it's" in "It means" but we have "it's" in "it's a signal" - we have that. We have "won't" - yes. We have "that's" - yes. We have no "" etc. We end with "The next test will be..." which is a concrete next step. Now we need to output JSON. Title: "Cathie Wood: Open-Weight AI Models Will Make OpenAI and Anthropic Richer" - but that's a bit long. We can shorten. We'll use: "Cathie Wood Says Open-Weight AI Models Enrich OpenAI and Anthropic" - that's good. Or "Cathie Wood: Open-Weight AI
