10^120 and 10^44: Musk, Chess.com, and the Game No Machine Can Solve

Last weekend, while rewatching game six of the 2026 World Championship match between Magnus Carlsen and Ian Nepomniachtchi, I suddenly noticed something that statistics never capture. The game stretched nearly eight hours, ending in an endgame from which most spectators had already left the arena by move ninety. On the analysis screen, engines kept producing evaluations that diverged by several percentage points. That was the moment I understood that even when machines have surpassed humans, a grey zone remains that no algorithm can erase. At the same time, on X, another debate was unfolding. Elon Musk — founder of xAI, a man who has publicly called chess "too simple compared to real life" — declared that with the evolution of superintelligence, chess would soon be completely solved. Chess.com, the largest online chess platform in the world with over one hundred million accounts, replied with a short, technically grounded line. The exchange spread quickly, drawing tens of thousands of interactions and becoming the most discussed topic in the chess community over the past week. On the surface this looks like nothing more than a social-media spat. But for me, someone who has spent thirteen years living with chess from the stands to the analyst's chair, it touches a more fundamental question: what are we misunderstanding about the complexity of chess — and what does that mean for how we speak about the limits of artificial intelligence? That V-League summer I learned something: a formation only looks beautiful when the opponent agrees to stand still. And in chess, the opponent never stands still. Two numbers, two worlds When people talk about the complexity of chess, they usually conflate two different concepts. The first is position space — the total number of legal positions that can exist on the board. Computed and refined by mathematicians over decades, it is now widely accepted to be roughly 10^44. The second is game-tree complexity — the total number of possible move sequences, including order and every branch. That figure stands at roughly 10^120, larger than the total number of atoms in the observable universe (about 10^80) by more than thirty orders of magnitude. Musk cites 10^44. Chess.com cites 10^120. Both are mathematically correct. But they are talking about two different things — and what decides whether chess can be "solved" is not which number looks better, but which one reflects the actual nature of the problem. A game is defined as "solved" when there exists a perfect strategy that determines the exact outcome — win, draw or loss — from every position, assuming both sides play optimally. Checkers was solved in 2026 after nearly twenty years of distributed computation, largely because its game tree "only" spans roughly 10^31 moves. Chess's tree is dozens of orders of magnitude larger. That is why chess has not been solved — and, by most current professional estimates, will not be solved in the near future, no matter how much computing power grows. But the story does not end there. And this is where I think both Musk and Chess.com fell short. The difference between "superhuman play" and "complete solution" What this exchange overlooks is a distinction any professional player understands clearly: modern AI does not "solve" chess. It plays chess at a level far beyond humans. These are two concepts on entirely different layers. AlphaZero, unveiled by DeepMind in 2026, learned chess from scratch in four hours and defeated Stockfish — the strongest engine of its time — in a hundred-game match. This milestone convinced many that chess had been "solved". In reality, AlphaZero made no claim about the optimal outcome of the starting position. It plays better than anyone, but it does not tell us whether chess is a win for White, a win for Black, or a draw under perfect play. I once fell into this trap. In 2026, when the pandemic halted world football, I was assigned to build a win-probability model for the first ten rounds after the Bundesliga resumed. I used the previous three seasons of data and believed my model reflected the true nature of the matches. The result: teams with a pre-pandemic xG differential better than 1.5 won only four of their first ten games after the restart, twenty-three percent below the historical average. Management was sceptical. I had to admit I had read the data without reading the context. That lesson applies exactly to the chess story — AI may play better, but that is not proof the game has been solved. An engine can call a position "slightly better for White" without being able to state the final outcome with certainty. There are endgames solved by tablebases containing trillions of positions, yet those tables only cover endgames with at most seven pieces. Extending to ten pieces already exceeds the storage capacity of every existing system. This is the most concrete evidence of how vast the gap between "playing well" and "solving completely" really is. A heat map can lie, but five consecutive failed tablebase extensions cannot. The blind spot of a tech billionaire Musk is neither the first nor the last person to use his tech credibility to speak about a field he does not practise. In 2026 he declared chess "too simple" because it has no fog of war, no tech tree, and both sides start from a symmetric position. This argument ignores one fact: precisely because there is no randomness, chess becomes the purest test of decision-making and calculation for humans and machines alike. The 2026 story is even more telling. That year, an internal chess event at Tesla was cancelled on Musk's instruction, on the grounds that chess did not serve the company's growth objectives. Details like these, pieced together, paint a clearer picture of his outlook: chess is a game to be overcome, not a domain to be understood. This is what actually matters. The Musk - Chess.com spat is not a debate between two technical positions. It is a dialogue between one side representing the chess community — with discipline, tradition, and deep understanding of the game — and another representing the AI industry, where tech prestige can be used to shape how the public understands the limits of intelligence. This is not the first such battle. But notably, this time, in an AI tournament running concurrently, Grok 4 — the model developed by Musk's own xAI — lost to OpenAI's o3. This is a detail mainstream media did not dwell on, yet it speaks volumes: even in the field where Musk claims AI will dominate, his own company is still not leading. I never claim that a billionaire has no right to speak about any field. But I do believe that the right to speak carries a responsibility proportionate to one's influence. When Musk makes a statement that hundreds of millions may read and believe, he is not speaking to the chess community — they know he is wrong. He is speaking to people who have never touched a chessboard, who may come to believe chess is a simple game waiting to be surpassed. What is actually at stake The question few ask is: if the public believes chess has been solved, what happens? First, that perception can affect sponsorship decisions. If chess is seen as a game with a known answer, the commercial value of major tournaments may decline. Not because chess is less exciting, but because the story it wants to tell — a clash of top minds — loses its drama. Second, this misperception can affect educational programmes that use chess as a tool for cognitive development. If parents believe chess is a game of the past, an arena with a known answer, they will invest less in their children's participation. This is a long-term risk chess organisations must proactively address. And third, there is opportunity. The Musk - Chess.com exchange pushed chess onto one of the largest media platforms on the planet. Chess.com seized the moment to reinforce its leadership in chess knowledge while reaching millions of potential new players. From a communications standpoint, this is a small but real win. The more worrying issue lies deeper. The way large tech organisations talk about the limits of AI is shaping how the public understands intelligence in general. When a figure of Musk's influence says that a domain as complex as chess is about to be fully conquered, he is not only talking about chess. He is building a mindset in which every complex problem can be reduced to a computing task — a mindset I believe is dangerous, both in chess and in life. What the next game will verify The Musk - Chess.com clash will probably fade in a few weeks. But the underlying question it raises will not disappear: how do we distinguish "the machine plays better" from "the machine has solved it entirely"? This is not a question for chess alone. It is the question every field — from football to medicine to education — will have to answer in the coming decade, as AI grows ever more capable of surpassing humans in specific tasks. No tactic is ever old, only the way we read the game goes out of date. And the way Musk reads the game — reading data detached from context, reading numbers without reading essence — has been out of date for a long time. If you want to verify what I have just said, here is a simple test. Open Stockfish, set up any middlegame position, and see what the engine says. It will give you a move, along with an evaluation like "+0.35" or "mate in 47". But it will not tell you that chess has been solved. That is the gap no social-media spat can erase — and it is what I will keep watching, game after game.