Two financial stories published this week trace opposite paths for the world's largest AI labs: OpenAI is moving toward a public offering that could value the company above $1 trillion, even while projecting a roughly $14 billion loss for 2026 — while Anthropic reported its first-ever quarterly operating profit.
What changed
According to reporting from Fortune, CNBC, and Yahoo Finance, OpenAI is currently generating about $2 billion a month in revenue (roughly $25 billion annualized), but projects a loss of approximately $14 billion for 2026 alone, with cumulative losses estimated at up to $115 billion by 2029. Even so, the company is targeting a public listing as early as September, with a projected valuation above $1 trillion — a jump from the $852 billion it raised at in its last private round, in March.
Anthropic went the opposite direction: according to CNBC (August 15), preliminary second-quarter revenue topped $11.5 billion — more than double the first quarter — and the company posted positive adjusted operating income for the first time, though the figures are preliminary and could still change. Earlier projections from the company itself pointed to roughly $559 million in operating profit on $10.9 billion in revenue, a margin of about 5%, driven by falling compute cost per revenue dollar.
Why it matters
For anyone evaluating or already using models from these two vendors, the difference between "burning cash to bet on future dominance" and "starting to operate in the black" isn't just a financial-market curiosity — it helps explain the pricing volatility we've covered here this same week, with DeepSeek's repricing and Gemini 3.7 Flash's promotional price with a built-in expiration date. Companies whose business model still depends on investor capital to sustain below-cost pricing have more reason — and more pressure — to reprice quickly than companies already covering their own operating costs.
The impact for Brazil
For Brazilian companies that depend on either of these vendors in production, vendor financial health should explicitly factor into long-term risk analysis — alongside security, governance, and performance. A vendor projecting billions in losses for several years running is making a bet that can shift course (price hikes, product discontinuation, changed priorities) more abruptly than a vendor with positive operating margin. That's not a reason to avoid OpenAI — it's a reason not to treat any vendor's current price, profitable or not, as guaranteed for the next few years.
Entercast's take
We've said before, covering DeepSeek's repricing and Gemini's expiring promotional price, that a listed price isn't a long-term constant. This week's financial contrast between OpenAI and Anthropic shows the structural root of that instability: part of the AI sector still runs on pricing subsidized by venture capital, not by its own revenue. When building the business case for an AI project, it's worth explicitly asking a question few teams ask today: if this vendor needs to balance its books within the next 24 months, what changes in price, support, or product continuity for what my company is using?