History Rhymes Again: Rise of Open-Source AI Models Incites Vitriol

The past, periodically staggering into the present like a shambling zombie, is never fully buried. How else are we to explain the recent social-media salvo from Dean Ball, OpenAI’s head of strategic futures, who channeled the spirit of former Microsoft CEO Steve Ballmer in a full-throated denunciation of open-source software.

We’ll get to Ball’s commentary shortly, but first I should explain the reference to Ballmer.

First, I want to clarify that Ballmer is still alive, currently (among other things) the owner of the NBA’s Los Angeles Clippers. Back in 2001, when he was in the big chair at Microsoft, Ballmer voiced an ambivalent opinion on open-source software. On the subject of open-source Linux, Ballmer actually leaped over the fence, charging into no-holds-barred animosity. Ballmer called Linux two c-words, namely “a cancer” and “communist.”

An article at ZD-NET, published a decade ago, recounts Ballmer’s Linux tirade. Quote:

“ . . . shortly after taking the helm at Microsoft, Ballmer said, "Linux is a cancer that attaches itself in an intellectual property sense to everything it touches."
The other c-word he used for Linux was communism, accusing the Linux community of diluting Microsoft's R&D spend by infringing on its intellectual property.
Reuters reports that Ballmer still stands by his cancer analogy as right for the time. Speaking at an event earlier this week, he said going to war with open source "made a ton of money" that still contributes to Microsoft's revenue. But he said he now considers that the threat from Linux is over.

That was then. Let’s move to the present.

From Past to Present

With the passage of time, perhaps OpenAI’s Ball will, like Ballmer before him, reach a similar point of contemplative revisionism. At the moment, however, he’s apoplectic — not about Linux, of course, but about the rise of China’s open-source LLMs, including Moonshot AI’s Kimi K3 model and Alibaba’s Qwen 3.8 Max. Ball features in a Wall Street Journal report that addresses Silicon Valley’s unease regarding the growing popularity of China’s open-source, open-weight models. Unlike proprietary LLMs from OpenAI and Anthropic (among others), open-weight models can be downloaded, adapted, and customized by enterprises.

In the Wall Street Journal article, Ball is quoted as follows:

“One probable outcome of an open-weight-model-dominant world is full AI communism, which is precisely what China proposes: rather than a market product, AI is a ‘public good’ which will ultimately be provided by the state as a kind of ‘digital public infrastructure,’” Dean Ball, OpenAI’s head of strategic futures, said in an X post Friday.
Ball, a former Trump administration official, called such a scenario a “dystopian hellscape” and said he expected the Trump administration to make moves toward reducing the use of open-source models. Ball later clarified that he wasn’t advocating for the U.S. government to discourage Chinese AI.

The Symmetry of Censure

It’s not entirely clear what Ball is advocating, but the stridency of his denunciation is remarkable, even in these unhinged times.

Still, you can see the symmetry from Ballmer to Ball. Even the name Ballmer is a prefixation of Ball, not that they’re related in anything other than their hostility toward variants of open source. On the enmity scale, Ball arguably outdoes his predecessor, not only playing the “communist” card, but reaching for the armageddon button of “dystopian hellscape.” It doesn’t get any more joyless than that, does it? I hope we never find out.

Not everybody, even at his current employer, is in agreement with Ball, who started in his current role just last month. The Wall Street Journal notes that “OpenAI have said his comments don’t reflect the company’s official positions.” This raises an obvious question: why have a head of strategic futures who speaks neither to the company’s strategy nor its future? Indeed, OpenAI has developed its own open models, and CEO Sam Altman now says, with the invaluable benefit of hindsight, that OpenAI’s approach of exclusively developing closed models was unwise.

Ball’s harsh rhetoric might be a genuine expression of his attitudes and beliefs. Then again, perhaps he gives the game away elsewhere in the article, where he mentions that large AI companies, presumably including OpenAI, raise billions of dollars to cover the expenditures of datacenters, processing infrastructure, and enhancements to AI models. To quote the Wall Street Journal, paraphrasing Ball: “If everyone uses AI systems that people largely don’t pay for, there would be no way to finance continued frontier AI development.”

I’m not sure that logic holds, especially if open-source, open-weight models — not only from China — were to redefine the cost and efficiency metrics associated with running LLMs at productive scale.

Is the Door to Change Ajar?

Despite the opposition to open-source models at Anthropic and (apparently to a lesser extent) at OpenAI, the Wall Street Journal article makes two salient points: first, that the use of Chinese open-source models is “surging” at U.S. companies, causing investors to question the viability of extravagant infrastructure expenditures by OpenAI and Anthropic; second, that purveyors of U.S.-based open-source models have emerged and are closing the performance gap on their Chinese counterparts. Vendors of open-source, open-weight models in the U.S. include Thinking Machines Lab, Nvidia (with its Nemotron 3 Ultra), and Reflection AI, which is backed by Nvidia.

The preceding sentence prompts a question regarding Nvidia. Is there a significant piece of the AI value chain that the company is willing to leave to other members of the ecosystem? Nvidia is already a leader in GPUs and accelerators, is moving into CPUs, is active on the “neocloud” front, sells network infrastructure that accompany its AI processors, and is getting into the LLM game, directly and indirectly. I suppose another question is whether these moves into what former Cisco CEO John Chambers termed “market adjacencies” are driven by offensive or defensive considerations? Yes, you might reply.

At any rate, if OpenAI’s Ball was attempting to curry favor within the Trump Administration, he wasn’t entirely successful. David Sacks, a prominent venture capitalist who serves as a White House AI adviser, interpreted Ball’s comments, which took the form of a post on X, as indicative of a “regulatory capture strategy.” In his own X post on Sunday, Sacks said: “The weaponization of regulatory uncertainty as a competitive tool should be completely unacceptable.”

Others in the White House might have other ideas, as the Wall Street Journal suggests. The debate apparently rages, in Washington and elsewhere.

Regardless, it’s not that difficult to envision a next wave of AI that will be less costly and more resource-efficient than what is offered by the LLM behemoths. As Ballmer looked back from 2016 on his intemperate words of 2001, he chose to contextualize and modify his verdict. Maybe, assuming a global rise of open-weight models worldwide (and not just from China), Ball will eventually adjust his views.