AI TRENDS | Can Cheaper AI Offset Data Center Construction Bottlenecks?

After U.S. voters elected a new slate of public officials this week, some technology executives and investors worried that delays in approving new computing facilities could raise the operating costs of existing AI servers and slow the ongoing decline in AI costs. According to Sina Finance, research firm Artificial Analysis found that OpenAI’s GPT-6.1-Sol, released in late September, matched the performance of the larger flagship GPT-6-Astra in high-reasoning mode while costing one-fifth as much to complete the same tasks. Anthropic has also promoted similar cost improvements in its latest model, even as Nvidia hardware purchase and rental prices have risen this year. As a previous report noted, falling prices need to continue or customers may reduce AI spending. A senior executive at a private equity firm that invests in and develops data centers said this week that advances in software and hardware would keep driving down model prices through token efficiency, or completing more tasks with less computing power. Those efficiencies could offset constraints on data center expansion, though the firm has invested in the sector and the executive has an incentive to express optimism. At an industry event hosted by Blackstone on Wednesday, Amazon Web Services CEO Matt Garman said he remained optimistic about the company’s data center expansion despite uncertainty from the election. “We are in a relatively favorable position, but I think other peers may face more challenges.” Amazon has faced strong local opposition to many data center projects and abandoned one in Virginia this year. Garman did not say whether server rental prices would continue rising. On Wednesday, shares of major cloud providers rose, while smaller providers with new facilities awaiting approval were sold off. A broader market pullback on Thursday erased some of the leading cloud providers’ gains. Most data centers scheduled to begin operating next year have already received all or some state and local approvals, so election-related policy backlash is expected to mainly affect computing projects planned for 2028 and later. Anthropic’s gross margin was about 85% in the first nine months of this year, excluding the cost of developing the models themselves. That leaves the company room to lower product prices if it needs to stimulate demand.