More than 10 basis points: that's how much long-term U.S. Treasury yields fell on average after major AI model releases, MIT Sloan researchers found. Isaiah Andrews and Maryam Farboodi examined 15 model-release events from five labs between January 2023 and December 2024 and found declines in Treasury yields, TIPS and corporate bonds that persisted roughly 15 trading days. Households and corporate borrowers feel the effect because the yield curve sets mortgage and corporate borrowing costs and signals investors' expectations for future consumption and growth. Yet market forces have pushed yields higher in 2026, with 30-year Treasurys trading above 5% in early May and the 10-year near 4.59% since the start of May, Business Insider reported alongside April 2026 CPI at 3.8% year-over-year.

The read here is direct. When major AI models land, the bond market nudges long-term yields lower, not higher. MIT Sloan's event-study found that long-term Treasury yields fell by more than 10 basis points on average after 15 model releases from January 2023 through December 2024. The researchers framed the result this way: "If market participants think that AI may have large growth effects, then new information about the direction of AI should impact long-term asset prices," they wrote. The drop lasted through about 15 trading days, suggesting the moves were not blips.

What the study measured and why it matters

Andrews and Farboodi tracked market moves around releases from five labs. They didn't limit the analysis to nominal Treasurys. Yields on inflation-protected securities, TIPS, and corporate bonds moved too. That scope matters because it ties the market response to expectations for real consumption growth and labor income, not just headline rates. A fall in long-term yields compresses mortgage rates and raises the present value of future cash flows, so the signal reaches households, corporate borrowers and financial markets.

Put plainly, market participants appear to treat model releases as forward-looking information about technology-driven growth. If new models change the expected path of productivity or labor income, bond investors will reprice long-term securities. The MIT Sloan paper interprets the consistent downward moves as evidence that investors update those expectations when they see concrete advances in AI.

Two channels: capex boost versus commoditization

That reading sits beside a contrasting channel Wellington Management has emphasized. Many economists expect an AI infrastructure buildout to raise GDP. Wellington cites estimates that AI-related capital spending could add roughly 0.75% to 1.5% to U.S. GDP soon. On its face, that would push long-term yields higher by lifting expected growth and inflation.

Wellington also points to an offset. The company DeepSeek has claimed top-tier models can be produced at far lower cost and plans to open-source its model. If those claims hold, model commoditization would lower the need for costly, ongoing infrastructure builds.

Mastheads

That would temper the capex impulse and could leave the downward pressure on long-term rates that the MIT Sloan event-study documents.

The two channels are simple. One raises yields by scaling up spending and pushing up expected demand. The other lowers yields by compressing future growth in capex if model costs fall and access broadens. The bond market's historical reaction to actual model releases, as measured by Andrews and Farboodi, suggests investors either expect the second channel to matter or are reweighting the two effects when new information arrives.

Those technical links are practical. A sustained reduction in long-term yields lowers mortgage rates, corporate borrowing costs and discount rates used by investors.

That changes project economics and household budgets. The yield curve is thus both a price and a signal.

Still, model-driven declines in long-term yields haven't been the only force recently. Business Insider reported upward pressure on yields in 2026. Thirty-year Treasurys traded above 5% in early May 2026, and 10-year yields rose to about 4.59% since the start of May, the outlet reported. Business Insider also cited April 2026 consumer prices at 3.8% year-over-year.

Two near-term dynamics help explain the divergence. Morningstar strategist Dominic Pappalardo warned that sustained energy-price inflation tied to the U.S.-Iran conflict and reduced Treasury purchases under incoming leadership could lift yields further, Business Insider reported. CNN's market coverage from December 2025 added that rapid Fed rate cuts in late 2025 removed a prior tailwind for lower bond yields, and expectations that the Fed will be cautious about further easing have left yields higher than they might have been earlier.

The picture is therefore mixed. The MIT Sloan event-study captures a reproducible, model-linked downward response in long-term securities that lasts about 15 trading days. At the same time, geopolitical risks, higher inflation prints and shifting central bank behavior have acted to push yields up in 2026.

My read is that investors are folding new technical information about AI into a market already contending with macro and geopolitical pressures. The net effect on yields will depend on which forces dominate over time, and whether model development follows a costly infrastructure path or a commoditization path like the one DeepSeek claims.

Related Articles

The concrete, dated finding from MIT Sloan is simple: long-term Treasury yields fell by more than 10 basis points on average after 15 major AI model releases, and that decline persisted for roughly 15 trading days.

This article was created with AI assistance.