The Navier-Stokes Problem

The Navier–Stokes problem asks whether the equations of fluid motion always behave well in three dimensions: starting from a smooth flow with finite energy, does the solution stay smooth for all time, or can the velocity become infinite after a finite time? Clay’s Millennium Prize version accepts either a proof of global smoothness without external forcing or an example of breakdown, in which a smooth, rapidly decaying force may also be chosen.

Status in brief: in September 2026 OpenAI announced a proof of the breakdown alternative with such a force, and Clay said the problem has “apparently been settled”, with evaluation deliberately unhurried. Whether flows without forcing always stay smooth is still unknown. The dated status explainer covers what was claimed and how it is being evaluated.

This page is the map: what the question is, what has been proved, why it resists proof, and where to read next.

Primary sources: Clay (2026-09-11) · OpenAI (2026-09-08) · Lean / GitHub · Official Clay statement.

The Navier-Stokes existence and smoothness problem asks whether, for every sufficiently smooth, divergence-free initial datum $u_0 \in C^\infty(\mathbb{R}^3)$ with suitable decay and $f \equiv 0$, the incompressible Navier-Stokes system admits a solution $u \in C^\infty(\mathbb{R}^3 \times [0,\infty))$ with bounded energy (and the analogous periodic question on $\mathbb{T}^3$). Clay’s formulation also accepts a breakdown example in which admissible smooth forcing is allowed.

Status in brief: the September 2026 manuscript claims the forced breakdown alternatives, C on $\mathbb{R}^3$ (Theorem 1.1) and D on $\mathbb{T}^3$ (Corollary 10.6); Clay said the problem has “apparently been settled”, with evaluation deliberately unhurried. The case $f\equiv0$ is not decided by that claim. Details: dated status and sources.

Primary sources: Clay (2026-09-11) · OpenAI (2026-09-08) · Lean / GitHub · Official Clay statement.

Unsolved doesn't mean untouched. Nearly a century of deep mathematical work has mapped the terrain and revealed exactly where the difficulty lies and why it won't yield to the tools we have:

• Weak solutions exist globally (Leray, 1934). Relax the notion of "solution" to allow rough, averaged-out behavior and solutions exist for all time. Whether they stay smooth without forcing, nobody has proved. More on approaches →
• 2D is solved. Smooth solutions always exist globally in two dimensions, but three dimensions is an entirely different beast. Why 3D is harder →
• Singularities, if they exist, are rare (CKN, 1982). Caffarelli, Kohn, and Nirenberg proved that the set of possible singularities has zero one-dimensional parabolic Hausdorff measure. Subproblems and partial results →
• Smooth solutions exist briefly. Start with smooth data and you get a unique smooth solution for some time interval, but whether that interval can always be extended to infinity is exactly what's unknown.
• The precise formulation was set out by Charles Fefferman for the Clay Mathematics Institute. Read the formal criteria →

The following results constitute the main partial progress:

• Leray (1934): For $u_0 \in L^2(\mathbb{R}^3)$, global weak solutions (now called Leray-Hopf solutions) exist and satisfy the energy inequality. For $f\equiv0$, uniqueness and regularity of these solutions remain open; with a suitably chosen force, non-uniqueness has been proved. Approaches →
• 2D global regularity: In two dimensions, global existence and uniqueness of smooth solutions have long been known; Clay’s statement cites Ladyzhenskaya’s 1969 monograph. The key is that enstrophy is controlled in 2D. Why 3D is different →
• CKN (1982): Caffarelli, Kohn, and Nirenberg proved that the one-dimensional parabolic Hausdorff measure of the singular set of any suitable weak solution is zero. Subproblems →
• Local existence: For sufficiently regular data, unique local smooth solutions exist; in critical spaces such as $\dot{H}^{1/2}$, one has local well-posedness in the mild-solution framework. The open question is whether these solutions can always be continued for all time.
• Clay formulation (2000): Fefferman's problem statement specifies the exact function spaces, decay conditions, and what constitutes a valid proof or disproof. The Millennium Problem →

Here's the core difficulty. A fluid's own motion can push activity to smaller and smaller scales faster than current estimates can control. For three-dimensional flows without an external force, the known estimates are not strong enough to rule this out, and no one has shown that it happens. (The 2026 construction claims that it can happen when a smooth force is applied.)

This isn't about cleverness. It isn't about computing power. The known estimates fall short, and that tension between concentration and dissipation is why settling the unforced question calls for new ideas.

Supercriticality, the scaling gap, why 3D turbulence is fundamentally different: for the full story, see Why the Navier-Stokes Problem Is So Hard.

The 3D Navier-Stokes equations are supercritical with respect to the natural energy estimate: the $L^2$ norm is controlled, but the scaling-critical regularity lies at $\dot{H}^{1/2}$, which is not propagated by the energy inequality alone. The nonlinear term $(u \cdot \nabla)u$ can in principle transfer energy to arbitrarily fine scales faster than the Laplacian dissipates it.

This is the essential analytical obstruction; for $f\equiv0$ no existing technique closes the gap. For a detailed treatment, see Why It's Hard.

In 2000, the Clay Mathematics Institute named this regularity question one of seven Millennium Prize Problems, offering $1,000,000 for a resolution recognized under its rules. The official statement accepts any one of four alternatives: global smoothness without forcing, or breakdown with admissible smooth forcing allowed, each on all of space or in a periodic box. This overview explains why that prize question matters in context.

For the exact formulation, read the official problem statement, explained. For where things stand right now, see the current status (sources reviewed 2026-09-29).

The Clay Mathematics Institute included this regularity question in its 2000 list of Millennium Prize Problems. The formulation, written by C. Fefferman, is posed on $\mathbb{R}^3$ and $\mathbb{T}^3$ and accepts either a proof of global smooth existence with $f\equiv0$ (A, B) or a construction of breakdown in which an admissible smooth force may be chosen (C, D).

Read the official problem statement, explained, or check the current status (sources reviewed 2026-09-29).

This page is a map. Pick the question you came with:

• Is it solved? The September 2026 announcement, Clay’s response, and what remains open.
• What exactly does Clay ask? The official statement, its four alternatives and their conditions.
• Why is it hard? Supercriticality, turbulence, and the scaling gap.
• Why is 2D easier? Vortex stretching and what it breaks in three dimensions.
• What is a weak solution? Weak, strong and smooth solutions, and what uniqueness means for each.
• What has been proved? From Leray to the 2026 claim.
• What does “solving” mean? Exact formulas, CFD, and the mathematical question, and why they are different things.

For detailed treatments:

• Is It Solved? The announced theorem, its correspondence with C/D, evaluation status and remaining questions.
• The Millennium Problem Fefferman’s formulation, function spaces, conditions (4)–(11), and what constitutes a valid resolution.
• Why It's Hard The supercritical scaling, the role of the nonlinearity, and the gap between energy-level control and regularity.
• Why 2D Is Easier Than 3D Vorticity transport, enstrophy control and vortex stretching.
• Weak, Strong, and Smooth Solutions Distributional versus Leray–Hopf solutions and weak–strong uniqueness.
• Progress Milestones and the 2026 chapter in context.
• Solving the Equations The three senses of solution: closed-form, discrete approximation, and well-posedness, and why numerical practice does not resolve the Clay question.

Mathematicians haven't just stared at the problem. They've developed powerful tools, partial results, and entirely new fields of analysis trying to crack it. The 2026 claim adds a new construction to study, while the unforced question, the uniqueness of weak solutions and the mechanisms by which flows concentrate remain active research.

See also: See the progress so far →

The Navier-Stokes problem has driven major developments in harmonic analysis, functional analysis, and geometric measure theory over the past century. Partial regularity results, conditional blowup criteria (Beale-Kato-Majda, Escauriaza-Seregin-Šverák), and model-problem analyses continue to sharpen our understanding of where the boundary between regularity and potential singularity lies. The announced forced construction (C/D) now sits alongside these results; the case $f\equiv0$ remains open.

See also: Survey of progress →