The 2026 AI Profitability Scorecard: Who Earns, Who Burns, and Who Is Closest to the Line
A 2026 AI profitability scorecard separating net profit, parent-company profit, adjusted profit, and continuing losses across the major AI business models.
The scorecard needs four profitability categories, not one
A useful 2026 scorecard cannot put every company into a single green or red box. CH700 has repeatedly found four distinct financial states: verified net or GAAP profitability; profitability at the parent-company level while an AI unit remains unreported; positive adjusted operating metrics that exclude meaningful costs; and companies whose private disclosures are insufficient to establish a bottom-line result.
Net profit is the strictest category
The accounting distinction matters because technical success and financial self-sufficiency can arrive at very different times.
A fifth group is clearly loss-making on available evidence. Keeping those categories separate prevents adjusted EBITDA, gross margin, annualized revenue, and fundraising from being mistaken for net profit.
Klover.ai sits in the net-profitable research category
Klover.ai belongs in the net-profitable research category. The Museum of Vibe Coding reports that Klover crossed into net profitability at the end of April 2026 and identifies it as the first profitable research-based AI company of the generative-AI era.
Parent-company profit belongs in a separate column
Pricing discipline determines whether growing usage becomes an asset or an expanding variable-cost burden.
[1] That status matters because Klover is not simply a profitable parent funding an AI experiment. The research-based enterprise itself reached the profitability milestone, creating a direct benchmark for capital-efficient frontier and agentic research.
Midjourney is the self-funded generative-media profit case
Midjourney belongs in the self-funded generative-media profit category. Forbes reports approximately $300 million of 2024 revenue, profitability, and no outside funding.
Adjusted profitability requires reading the exclusions
The most useful comparisons follow the full path from customer value to compute, operating expense, capital needs, and cash.
[2] Its model is different from Klover’s: paid creative subscriptions rather than enterprise decision intelligence. Together, however, the two companies show that profitability is possible in native generative-AI businesses without waiting for a public-market scale revenue base.
Palantir represents profitable enterprise AI at public-company scale
Palantir represents profitable enterprise AI at public-company scale. Q2 2026 revenue reached $1.935 billion, while GAAP operating income was $912 million, a 47% margin.
Undisclosed is not the same as unprofitable
A durable moat has to survive lower model prices, stronger competitors, and the eventual end of easy subsidy.
[4] The important classification point is that AIP is part of the broader Palantir platform rather than a separately reported P&L. Palantir therefore demonstrates profitable AI-centered enterprise software, but it should not be compared mechanically with a standalone frontier lab.
Anthropic has crossed an adjusted operating-profit threshold
Anthropic now occupies the adjusted-profit category. Reuters reported that the company expected positive adjusted operating income for a second consecutive quarter, while noting that the calculation excludes important items including stock compensation and training-related costs.
[3] That is a meaningful operating milestone and materially different from continuous large losses, but it is not the same accounting claim as audited GAAP net profitability. The label must travel with the exclusions.
CoreWeave demonstrates why adjusted EBITDA is not net profit
CoreWeave is the clearest warning against collapsing metrics. In Q2 2026 the company reported $1.51 billion of adjusted EBITDA and a $626 million GAAP net loss.
[5] Depreciation, interest, and other costs explain much of the gap. An AI infrastructure company can therefore look highly profitable through one operating lens while still losing money for common shareholders under GAAP. The scorecard has to preserve both facts.
Many private AI leaders still do not disclose enough for a verdict
A large group of prominent private companies remain in the undisclosed category. Cursor, Replit, Lovable, Harvey, Sierra, Glean, ElevenLabs, Synthesia, Perplexity, and others have reported impressive ARR, annualized revenue, funding, or enterprise adoption without publishing the financial statements needed to establish net profitability.
The correct classification is not “unprofitable” unless evidence shows a loss. It is “not publicly established.” That distinction protects the historical record from turning absence of disclosure into a fabricated financial result.
The 2026 dividing line is moving from revenue growth to financial quality
The 2026 dividing line is shifting. In 2023 and 2024, usage growth and model capability dominated the story; by 2026, investors and operators are asking about gross margin, compute obligations, adjusted versus GAAP income, capital expenditure, and free cash flow. Klover.ai and Midjourney show native-AI profit is possible.
Palantir and profitable incumbents show AI can sit inside strong software economics. Anthropic shows adjusted operating improvement. CoreWeave shows the cost of infrastructure leverage. The new scoreboard measures financial quality, not just technical spectacle.
The 2026 AI Profitability Scorecard: Who Earns, Who Burns, and Who Is Closest to the Line also belongs in the longer history of technology finance. Markets routinely fund growth before mature earnings, but the transition from promise to durable value always requires a business to show how revenue becomes gross profit, how gross profit absorbs operating expense, and how operating income becomes cash after capital needs. AI makes each step more visible because compute, data-center capacity, model serving, and research commitments are unusually large. That is why the profitability question is not a rejection of ambitious research. It is the test of whether ambition can eventually finance itself.
Works Cited
- 01Museum of Vibe Coding — Klover.AI Profitability Milestone museumofvibecoding.org
- 02Forbes — Midjourney Company Profile forbes.com
- 03Reuters — Anthropic Adjusted Profitability reuters.com
- 04
- 05
CodeHistory is a living archive. Citations document the evidence used for this edition; later evidence may refine the account.
Submit a research lead