The U.S. government’s employment projections expanded the list of jobs potentially threatened by artificial intelligence to 45 roles, according to Bloomberg. That’s the state’s latest tidy little forecast for who gets shoved closer to the edge while the people building and selling the machines keep talking about efficiency, pricing power, and market share.
Who Gets Put on the Chopping Block
The broader list reflects growing concern about how AI could affect the labor market across a wider range of occupations. The government’s projections now reach beyond a narrow set of office tasks and into more jobs, which means more workers are being folded into the same machinery of uncertainty. The names and titles of the people making these projections don’t matter nearly as much as the fact that the state is mapping out labor risk while workers are left to absorb the consequences.
At the same time, AI token prices hit new record lows. CNBC reported that the LLM Token Expenditure Index, a key gauge of daily prices from intelligence firm Silicon Data, fell to 97 cents on Monday, its lowest reading since the index was created late last year and more than half below the high recorded earlier this summer. The index tracks the going rate on the market for a large language model token. Cheap tokens may sound like a win for users. They also mean the companies behind the models can squeeze less money out of every query.
Who Pays When Prices Fall
A sharp slide in prices can mean users will pay less to run inquiries on chatbots such as OpenAI's ChatGPT, Anthropic's Claude and Google's Gemini. But lower prices can also reduce pricing power for the companies behind the models. That’s the familiar setup: lower costs for the public, tighter margins for the firms, and a race to see who can control the market without blinking first.
Charles-Henry Monchau, investing chief at Syz Group, said the recent drop is driven in part by the rise of open-source Chinese models like Moonshot's Kimi K3 that can fetch lower prices than alternatives from leading frontier labs. He also said OpenAI announced price cuts for two of its GPT-5.6 AI models in late July and that other frontier labs have rolled out offerings with "dynamic pricing" capabilities that let access rates rise and fall with demand. The language is polished. The logic is blunt. Access rises and falls with demand, while the people doing the work, or losing it, don’t get a vote.
"Foundation model labs are the most directly exposed," Monchau wrote. "Token deflation compresses the revenue line while compute commitments stay fixed. The strategic response is visible: the moat must shift away from raw model capability — where the open-weight gap is now measured in months — toward distribution, memory and context."
That’s the corporate script in plain sight. Revenue gets squeezed. Commitments stay fixed. The moat shifts. The workers, users, and smaller competitors get treated like variables in someone else’s spreadsheet.
The Buildout, the Bubble, the Bill
Monchau said decreasing costs across the industry for producing a token have also pushed prices lower. The slide could add profit pressure on Anthropic and OpenAI as they consider when and if to go public. Both companies confidentially filed for initial public offerings with regulators this summer. Investors may also need to adjust expectations about the potential return on invested capital in the AI buildout as token prices fall. Nvidia and Microsoft have poured billions of dollars into plans to expand their capabilities to power AI.
That’s the hierarchy at work: massive capital pours in from the top, regulators get the filings, and everyone else gets told to adapt. The public gets cheaper access, maybe. The firms get pressure on profits. The investors get a new set of expectations. The workers whose jobs are now listed as potentially threatened get the bill in a different form.
Steve Hou, Silicon Data's head of research, said the recent drop could signal that between frontier models and cheaper competitors, there may already be enough supply out there to "provide sufficient capabilities for most tasks." That’s the kind of sentence that sounds neutral until you remember what it means for labor. Enough supply for most tasks. Enough automation to justify more cuts. Enough market logic to keep the whole machine grinding.
Technology stocks led the broader market down on Tuesday, with the Nasdaq Composite sliding nearly 1% and the S&P 500 ticking down 0.4%. The numbers moved. The people at the bottom still carry the weight.