A new whitepaper from GQG Partners has issued a stark warning on OpenAI’s long-term business viability, arguing the company’s economics are fundamentally unsound despite rapid revenue growth, mass user adoption and its central role in the global AI infrastructure boom.
In Dotcom on Steroids: Part II, GQG contends that OpenAI’s business “appears commoditized” and is “hyper capital-intensive,” with rising compute costs and intensifying competition from both closed- and open-source developers.
“We believe it will struggle to build a sustainable business over time,” the paper stated.
GQG argued that open-source AI labs are now producing models comparable to leading closed-source systems, a shift it said will “lower switching costs for users and erode the stickiness and competitive advantage” OpenAI has relied upon.
The whitepaper warned that OpenAI is deeply exposed to capital-market sentiment because of its pivotal role in the AI buildout.
If investors begin to doubt that artificial general intelligence is achievable “within a reasonable timeframe”, the firm could face higher borrowing costs or reduced funding, with potentially “massive” spillover effects.
GQG described OpenAI’s financial profile as “broken and unrealistic” by any historical benchmark, even when judged as a high-growth start-up.
It highlighted the disconnect between the company’s valuation and its roughly US$20 billion revenue run rate, pointing to Amazon and Google’s significantly higher revenue levels when they reached similar valuations.
The firm argued OpenAI sits at the centre of “a web of circular financing” linking major public-market technology companies, contending that these relationships now have implications well beyond private markets.
It also said the company has helped fuel an AI bubble by promoting expectations of rapid progress toward artificial general intelligence, despite experts increasingly questioning that trajectory.
At an AI conference in Miami, OpenAI CFO Sarah Friar told investors and lenders the company’s “true north” is “AGI for the benefit of humanity.”
While GQG has acknowledged OpenAI’s achievements, including more than 800 million weekly ChatGPT users and strong enterprise adoption, it argued that this success has been underwritten by “record-breaking capital infusions and expenditures” on infrastructure, supported by what it describes as overly optimistic long-term forecasts.
A key concern is the durability and quality of demand, according to GQG, with the firm stating that a large share of OpenAI’s usage comes from free-tier users engaging for “non-productive reasons,” arguing this inflates activity metrics while offering limited insight into sustainable revenue.
GQG also pointed to reliability issues—especially “hallucinations”—that hinder mission-critical adoption.
Enterprise stickiness remains limited, according to the paper, with “28 per cent of OpenAI’s API usage” flowing through low-code platforms that can easily switch to cheaper or more efficient models.
GQG also raised concerns that benchmark tests may overstate real-world capability, arguing they reward breadth rather than the consistency enterprises require.
The firm stated that although token prices have fallen, the cost per query has not meaningfully declined due to rising complexity, verbosity, repeated runs to suppress hallucinations and multi-step reasoning workflows that dramatically increase token consumption.
GQG questioned how much further generative AI can improve, noting slowing benchmark gains and sharply rising training costs, and argued that LLMs face “hard constraints,” including limited high-quality human data and risks associated with synthetic data.
The paper outlined that OpenAI’s business model lacks the scalable economics of software because compute costs scale directly with usage, preventing meaningful economies of scale or network effects and questioned subsidised pricing strategies in highly price-sensitive markets such as India.
While enterprise spending on AI foundation models has more than doubled, GQG cites studies showing many companies remain stuck in pilot phases or achieve modest productivity gains.
One study estimated AI may improve profit margins by just 50–70 basis points by 2030.
OpenAI generated US$4.3 billion in revenue in the first half of 2025 but reported a net loss of US$13.5 billion and a cash burn of US$2.5 billion.
It forecast profitability by 2030 with revenue of US$200 billion and gross margins above 60 per cent—projections GQG likened to “dotcom-era” extrapolations.
GQG highlighted intensifying competition from major labs and fast-moving open-source developers.
It pointed to China’s open-source “Kimi K2” model, which has matched or exceeded leading closed-source reasoning systems, arguing that “good enough” open-source models could become industry standards, similar to how Linux underpinned the rise of Android.
The firm also warned of talent-retention risks, noting OpenAI is on track to spend nearly US$6 billion on stock-based compensation in 2025 tied to a US$500 billion valuation.
GQG said OpenAI has raised more than US$40 billion this year, up from US$6.6 billion in 2024, and may need to continue raising capital at unprecedented levels to subsidise compute, retain talent and sustain operations.
The paper referenced past board concerns, noting court filings and accounts showing OpenAI’s board removed Sam Altman in 2023 for not being “consistently candid in his communications.”
GQG concluded that while large language models may be enduring, the economics underpinning their development may not be.
“We believe the story to watch is not whether the technology is immortal,” GQG stated, “but whether the companies building it are.”
