OpenAI’s $12.3B quarterly loss exposes a new AI profit split

A new scoreboard for frontier AI

The latest private-market scorecard for frontier AI no longer reads simply as OpenAI first and everyone else chasing. Fresh reports say OpenAI generated $6.7 billion in second-quarter revenue, up 18% from the first quarter, but its operating loss widened to $12.3 billion from $9.3 billion [1]. Anthropic, by contrast, reported more than $11.5 billion in second-quarter revenue and positive adjusted operating income, making the financial comparison between the two largest independent AI labs far more complicated than a model-performance contest [6].

The headline is dramatic, but the fine print matters. OpenAI’s loss figure is described as an operating loss that includes stock-based compensation, while Anthropic’s profitability is framed as adjusted operating income, a metric that is not directly comparable because it excludes some costs [6]. That means “Anthropic profits” should not be read as a clean claim that the company has proven durable GAAP profitability; it means the Claude maker is showing a path to positive operating economics under an adjusted measure while OpenAI’s spending still overwhelms its revenue [6].

Still, the direction of travel is striking. OpenAI is no longer merely a high-growth company investing ahead of revenue; it is a company whose reported quarterly loss is nearly double its quarterly revenue [1]. Anthropic, meanwhile, is using its revenue surge and adjusted profit story to support an IPO narrative that could arrive before OpenAI reaches the public market [5].

What changed in the quarter

The second quarter appears to have marked the first time Anthropic overtook OpenAI in quarterly revenue, with reported sales above $11.5 billion compared with OpenAI’s $6.7 billion [6]. TipRanks reported that Anthropic’s sales rose from $4.73 billion in the first quarter to more than $11.5 billion in the second, a gain of about 143% quarter over quarter [6]. The same report said Anthropic’s adjusted operating profit came partly from more efficient use of computing resources, while also warning that the company’s metric excludes stock-based pay [6].

OpenAI’s numbers point in the other direction. Its revenue grew from $5.7 billion in the first quarter to $6.7 billion in the second, but the operating loss increased faster, rising 32% to $12.3 billion [6]. That gap is central to the current investor debate: OpenAI is producing large and still-expanding revenue, but each additional dollar appears tied to heavy compute, product and user-acquisition costs [1].

The broader market context is also changing quickly. Lapaas Voice reported that, after launching GPT-5.6 Sol on July 9, OpenAI’s revenue rose 35% this quarter and enterprise revenue grew by more than 50% [3]. It also cited Ramp data showing OpenAI business API spending up 82% quarter over quarter, compared with 76% for Anthropic, suggesting OpenAI may have regained momentum in business usage even after losing the second-quarter revenue comparison [3].

That makes the current picture less like a clean victory for one side than a split screen. Anthropic has the cleaner second-quarter profitability narrative and the larger reported quarterly revenue base; OpenAI has signs of a faster third-quarter commercial rebound after a new flagship model release [3].

Why the loss is so large

The immediate reason for OpenAI’s $12.3 billion loss is the cost of building, training and serving frontier AI at global scale. Fresh reporting ties the loss to data-center and computing costs, which remain the central expense category for companies that must train large models and then run them for hundreds of millions of users or enterprise workflows [1]. OpenAI also carries the economics of a mass-market consumer product: ChatGPT’s free and lower-priced users help preserve scale and brand power, but they can also increase inference costs without generating proportional revenue.

The subscription model may not solve the problem if heavy users consume far more than they pay. DigitalToday reported that analysis of premium AI subscriptions found roughly $200-a-month plans can deliver API-equivalent usage worth thousands of dollars when users run long coding tasks, including estimates of about $14,000 in API use for OpenAI and $8,000 for Anthropic [4]. The same report said agent-based coding can consume around 1,000 times more tokens than simple coding questions, turning high-engagement customers into potential cost centers if pricing and usage limits are too generous [4].

This is the core economic tension for generative AI. A product can be beloved, sticky and fast-growing, yet still lose money if usage expands faster than monetization. The most valuable customers are enterprises that embed models into workflows and pay recurring API or subscription fees, but the same enterprise workflows can produce enormous token volumes when agents perform multistep tasks [4].

Anthropic’s advantage, and its caveat

Anthropic’s current advantage is not simply that it grew faster; it is that its growth appears more directly tied to enterprise demand and coding workflows. The company’s Claude products have been gaining traction among developers and corporate users, and current reporting says Anthropic crossed an annualized revenue run rate above $65 billion while posting $11.6 billion in second-quarter revenue [4]. That is why investors are attaching extraordinary importance to the company’s IPO prospects and its ability to keep expanding compute supply [5].

But Anthropic’s profit narrative also deserves restraint. The reported profit is adjusted operating income, not necessarily full net income after all costs, and the adjustment issue is especially important in companies where employee equity compensation can be large [6]. In other words, Anthropic has shown that a frontier AI lab can report positive operating economics under a favorable measure, but it has not ended the debate over whether frontier model companies can generate durable shareholder returns after infrastructure, dilution and competition are fully counted [6].

The competitive risk is also obvious. OpenAI’s post-Sol acceleration suggests the revenue race remains fluid [3]. Open-weight and cheaper models are pressuring pricing across the market, and DigitalToday reported that Anthropic’s most expensive model, Fable 5, accounted for only 6% of total tokens purchased by customers in its first month after launch [4]. If buyers can route simpler tasks to cheaper models and reserve frontier systems for only the hardest jobs, premium labs may face pricing pressure even as usage keeps rising [4].

IPO timing turns the numbers into a public-market test

The financial split matters because both companies are moving toward public-market scrutiny. Trades Prophet reported that OpenAI CFO Sarah Friar told employees the company plans to be public in 2027, or earlier if growth accelerates [1]. The same report said Anthropic filed its confidential prospectus in June and could list as early as September, potentially putting Claude’s parent in the public market ahead of OpenAI [1].

The contrast will be central to both roadshows. OpenAI can argue that it remains one of the fastest-growing technology companies, with a July annualized revenue run rate above $40 billion and a renewed enterprise push after GPT-5.6 Sol [1]. Anthropic can argue that it has crossed OpenAI on reported quarterly revenue and has already reached adjusted operating profitability [6]. Public investors will then ask which is more valuable: OpenAI’s consumer reach and platform ambition, or Anthropic’s narrower but apparently more monetizable enterprise mix.

Valuation makes the question sharper. Cryptopolitan reported that investors are discussing a possible valuation around $2 trillion for Anthropic’s planned IPO, while also noting public resistance to the AI data centers required to sustain its growth [5]. At that scale, “profitable this quarter” is not enough; investors would need confidence that margins can survive years of model competition, infrastructure expansion and pricing pressure.

The data-center overhang

Compute is now both the growth engine and the balance-sheet risk. Conservative Daily News reported that Big Tech’s AI buildout is increasingly tied to debt and local infrastructure constraints, with Texas pausing approvals for more than 1,800 grid-connection projects while officials review power, water and community impacts [2]. The same report said OpenAI’s $6.7 billion of second-quarter revenue came alongside the $12.3 billion operating loss, illustrating how rapid demand growth can coexist with worsening economics [2].

For Anthropic, the challenge is different but related. It needs enough data centers, chips and electricity to meet demand for Claude and newer AI products, and Cryptopolitan reported that local opposition to data centers is emerging as a risk likely to appear in IPO disclosures [5]. Data-center resistance does not necessarily mean AI demand is weak; it means supply can become slower, more expensive and more political [5].

That is why the OpenAI-Anthropic comparison is more than a rivalry story. It is a stress test for the whole AI business model. If the best-known consumer AI company can lose $12.3 billion in a quarter while the enterprise-focused challenger reports adjusted profit, investors will likely reward disciplined monetization over raw usage [6]. But if OpenAI’s third-quarter rebound continues and enterprise customers return to its API at faster rates, the market may decide that temporary losses are the cost of holding the broader platform [3].

The bottom line

The current evidence supports a nuanced conclusion: Anthropic appears to have won the latest reported quarter on revenue growth and adjusted operating profitability, while OpenAI remains a growth giant with a serious cost problem but renewed third-quarter momentum [3]. The most important distinction is not “AI works” versus “AI does not work”; it is whether each company can convert massive token consumption into margins that survive competition, infrastructure bottlenecks and public-market accounting [4].

For now, Anthropic has the cleaner IPO story. OpenAI has the bigger brand, a new enterprise acceleration narrative and a loss figure that investors cannot ignore [1]. The next decisive metric will not be a chatbot benchmark. It will be whether either company can make growth cheaper to serve.

Sources from the last 72 hours

1. [1]OpenAI’s CFO just said something IPO investors should hearAug 22, 2026, 12:00 AM UTC
2. [2]Is Big Tech Sowing Seeds Of Its Own Destruction?Aug 22, 2026, 12:00 AM UTC
3. [3]OpenAI Revenue Rises 35% After GPT-5.6 Sol Model LaunchAug 22, 2026, 12:00 AM UTC
4. [4]Paying $200 a month for $14,000 in use adds pressure on AI profitabilityAug 24, 2026, 11:00 AM UTC
5. [5]Anthropic’s $2 trillion IPO faces a new threat as Americans push back against AI data centersAug 22, 2026, 12:00 AM UTC
6. [6]OpenAI Bleeds $12.3 Billion in Q2 as Anthropic Overtakes With a 143% Revenue SurgeAug 22, 2026, 12:00 AM UTC

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.