Fluid equations can develop infinite velocity under sustained force

A computer proof showing that ideal fluids can tear themselves apart challenges mathematical assumptions while exposing the boundary between automated search and human understanding.

A smooth stream of water appears continuous and predictable, guided by internal friction that disperses motion before any single point can accelerate out of control. For nearly two centuries, the mathematical equations governing fluid mechanics have treated liquids and gases as continuous media, assuming that smooth beginnings guarantee smooth movement into the future. Yet beneath that practical assumption lies an unresolved theoretical question: whether the equations themselves might allow velocity to concentrate into an infinitely sharp point, causing the mathematical description of the fluid to tear itself apart.

Understanding whether such a breakdown occurs matters because differential equations form the primary language used to describe the physical world.¹² If the equations of fluid flow can generate an infinite velocity from well-behaved initial conditions, the mathematical continuum fails to hold. While physical liquids consist of discrete molecules that prevent actual infinities, the existence of a mathematical singularity reveals where continuum mechanics ceases to provide a consistent picture of flow.²¹

For a fluid to develop such a breakdown, specific physical processes must align in a precise cascade. A pocket of fluid must first begin to spin, drawing surrounding material into an inward spiral. As this vortex filament contracts, conservation laws force its rotational speed to climb, stretching the core axially. For the equations to blow up in finite time, the internal forces of pressure, acceleration, and viscous dissipation must balance each other almost entirely, allowing the velocity gradient at the very center to diverge toward infinity while total kinetic energy remains finite. If external energy enters the system, that push must remain mathematically smooth throughout, avoiding any artificial roughness that could force an unnatural failure.

Can smooth fluids tear themselves apart?

The Navier-Stokes equations can indeed develop an infinite velocity under continuous smooth forcing, according to an analytical proof announced on September 8, 2026, by researchers at OpenAI.²¹ In the official formulation framed in 2000 by Charles Fefferman of Princeton University for the Clay Mathematics Institute, the $1 million Millennium Prize problem offers four possible paths to resolution: proving that smooth solutions always exist without forcing, or proving that a finite-time breakdown, known as a blowup, occurs under smooth forcing.²¹

OpenAI reported that its proof resolves Statements C and D of the Clay Institute formulation by constructing a scenario in which an initially motionless fluid subjected to smooth forcing forms a singularity within finite time.³² The company stated that the constructed solution behaves like a vortex filament, spiraling inward while being stretched along its axis until velocity gradients diverge.²¹ However, OpenAI stated in its published write-up that it does not intend to claim the $1 million Millennium Prize for the result, focusing attention instead on whether its specific forcing function satisfies the strict decay conditions established by Fefferman.³²

How did artificial agents construct the proof?

Autonomous artificial intelligence systems derived the argument by searching mathematical solution spaces in parallel over several days. OpenAI reported that an unreleased internal model, described by the company as significantly more capable than its GPT-6 Astra system, powered the search.⁴² Across an 88-hour computational run from September 1 to September 5, up to 10,000 concurrent autonomous agents communicated through 2.7 million messages and generated approximately 130 billion output tokens to assemble the proof.³² A subsequent syntax verification of the formal proof required 17 hours using GPT-6 Astra in the interactive theorem-proving language Lean.³²

The computational achievement followed closely on the heels of parallel work by Tristan Buckmaster of New York University and Levent Alpöge at Anthropic.¹² Buckmaster and Alpöge announced Lean-verified proofs establishing finite-time blowup for the related three-dimensional Euler equations, which govern idealized fluids that have zero viscosity.¹² Both efforts built directly upon analytical layering techniques pioneered by Diego Córdoba and Luis Martínez-Zoroa at the Institute of Mathematical Sciences in Madrid, who had previously demonstrated singularities using rougher, non-smooth forcing terms.¹

What does the proof leave unresolved?

The result is a formal mathematical construction rather than an empirical measurement of flowing water, and it cannot describe the behavior of real molecular fluids. Because real liquids are composed of atoms, molecular kinetics take over long before a mathematical velocity gradient reaches infinity. Furthermore, formal verification in Lean confirms syntactic and deductive consistency within specified axioms, but Lean cannot judge whether the physical decay conditions of the forcing term fully satisfy the formal criteria established for the Millennium Prize.²¹

Questions also persist regarding the transparency of the search process and the conceptual clarity of machine-derived arguments. Buckmaster noted that early machine-generated proofs in his collaboration were exceptionally difficult to digest, observing in an announcement statement: “the first llm generated proof levent sent me was the most horrendous i have ever read.”⁴¹ Field Medalist Terence Tao of the University of California, Los Angeles, cautioned before the announcement that a proof derived through a black-box search provides little intrinsic value to mathematics if mathematicians cannot absorb its underlying conceptual insights.²

Thomas Wolf, an artificial intelligence researcher, argued in a public analysis that finding a counterexample resembles finding a needle in a haystack through massive search, leaving open whether automated systems understand mathematical elegance or fertile research directions.⁴ Whether the Navier-Stokes proof yields broader physical principles will depend on whether mathematicians can distill its thousands of lines of machine verification into ideas that human minds can understand.

This piece was prepared from research announcements and public records; the authors have not been interviewed.

References

This article is based on 4 sources, listed in the order they are cited.

1. 1 QM Quanta Magazine third party · 8 Sep 2026 AI Has Solved One of Math’s $1 Million Millennium Prize Problems | Quanta Magazine See the source
2. 2 X XenoSpectrum third party · 8 Sep 2026 OpenAI Claims Its AI Proved a Navier-Stokes Blowup — Mathematicians Question Priority and Verification | XenoSpectrum See the source
3. 3 T thenextweb.com third party · 9 Sep 2026 OpenAI publishes its Navier-Stokes proof and says it will not claim the Millennium Prize See the source
4. 4 H https://x.com/Thom_Wolf announcement · 9 Sep 2026 AI Model Produces Mathematical Proof With Minimal Human Input See the source