This piece lays out what was actually said on both sides, without taking a position on who “won” the exchange. Both Elon Musk and Chess.com made claims that turned out to be factually grounded, just about two different things.
On the evening of 3 September 2026, Elon Musk did what Elon Musk tends to do online. He replied to someone else’s post on X (formerly twitter) with a confident, sweeping claim, and by morning half the internet’s chess accounts had opinions about it. The post he replied to came from an account with the username @Rainmaker1973, a viral thread breaking down how quickly chess’s possible move sequences multiply. Musk wrote that chess has very few genuinely good moves and predicted the game would eventually be fully solved, the same way checkers has been.
Chess.com’s official account quote-tweeted him back with two words: “skill issue.”
Two days of public sparring followed, pulling in millions of views, and underneath the jokes sat a real question that chess mathematics had quietly settled long before either account existed.
How the Argument Unfolded
Musk followed up the next morning, arguing that “chess is hard for humans, but not for computers.” He added that people would keep enjoying the game anyway, the same way people enjoy sports where machines have already surpassed human performance. The exchange caught the attention of the wider chess world. Grandmaster Alireza Firouzja commented that he did not mind either way, so long as SpaceX shares kept climbing. YouTuber Levy Rozman, aka GothamChess, offered a dry “Haha yeah.”
But Chess.com then brought in a number. The account pointed out that the total count of possible chess games is roughly 40 orders of magnitude larger than the number of atoms in the observable universe. Musk pushed back, arguing that the relevant figure was not the number of possible games. He said the number that mattered was the count of legal board positions, which he put at around 40 orders of magnitude fewer than the same atom count. He went further, suggesting that sufficiently advanced future AI would eventually be capable of working through that entire problem in ways no human today could follow or verify. Both were describing real numbers, just not the same number.
At one point Musk dismissed the person running Chess.com’s account as an intern. The account pushed back, noting they were a full-time employee. However, this exchange moved on in good humour rather than turning genuinely hostile.
The Argument That Chess Will Be Solved Has Come Up Before
This is not the first time Musk has raised this idea. In 2022, he posted that he had stopped playing chess as a child, finding it too simple to matter in real life. Chess.com responded with a lengthy blog post defending the game’s complexity. The post took a pointed shot at the knight’s movement, calling it “not random, and probably confusing to someone who is a billionaire.”
In May 2024, Musk made a near-identical prediction, putting a 10-year timeline on chess being fully worked out by machines. It was a claim close in substance to what he posted this September. By October that year, he made a similar point again. He joked that the game would need something closer to a weapon capable of destroying a king from orbit before it could compete with how complicated real life actually is. Seen together, the 2026 exchange was not a new argument. It was at least the fourth time Musk has raised a version of the same idea.
But, Will Chess Really Be Solved?
In game theory, a game is solved when researchers have mathematically proven the outcome of perfect play from both sides, which means a guaranteed win for one side, or a guaranteed draw. That is a much narrower claim than “a computer can beat every human,” which chess engines have already achieved. Checkers reached the first kind of solved status in 2007, when researchers proved that perfect play always ends in a draw. Chess as a whole has never been solved that way, and nothing about the current generation of engines changes that.
A meaningful slice of chess has been solved already, though. Every endgame position with 7 or fewer pieces on the board, kings included, has been fully mapped by computer tablebases, complete since August of 2018. For any of those positions, the perfect result and the perfect move are already known with certainty. Work on eight-piece tablebases is still ongoing. What tablebases have not touched, and what nobody has come close to solving, is the full game, starting from the opening position with all 32 pieces still on the board.
Two Numbers, One Shared Origin
Both figures in the Musk-Chess.com exchange trace back to a single 1950 paper, “Programming a Computer for Playing Chess.” It was written by Claude Shannon, the founder of information theory, and it is widely credited as the paper that started the field of computer chess. Shannon was trying to prove that brute-force calculation could never solve chess; checking every possible game one at a time was hopeless. To make that case, he estimated the total number of possible chess games at around 10^120. That figure is now called the Shannon number, the same one Chess.com cited 76 years later.
Shannon also estimated the number of legal positions separately, arriving at a smaller figure of his own. Modern research by chess mathematician John Tromp has refined that count to roughly 4.6 x 10^44. That number is close to what Musk was describing, without naming where the concept came from. Chess.com was talking about total possible games. Musk was talking about legal positions. Shannon had already drawn that exact distinction himself, for the same reason Musk was reaching for it now.
Kasparov Made the Same Point in 2010
16 years before this elon musk vs chess.com thing, former World Champion Garry Kasparov wrote an essay for The New York Review of Books on chess and computers. He called chess “far too complex to be definitively solved with any technology we can conceive of.” He pointed to checkers being weakly solved by Jonathan Schaeffer’s program Chinook as the closest real precedent. He also used the same atoms-in-the-universe comparison that Chess.com would later use in this exchange.
What Musk’s Own AI Has Actually Managed So Far
Part of Musk’s argument rested on the idea that future artificial intelligence would eventually solve chess in ways far beyond human comprehension. That claim is worth checking against a real result. In August 2025, xAI’s Grok 4 competed against OpenAI’s o3 model in the Kaggle Game Arena AI Chess Exhibition. The tournament barred chess engines and specialised training, leaving each chatbot to rely only on general knowledge picked up from the internet. OpenAI’s o3 won all four games.
World number one Magnus Carlsen, currently playing in the Global Chess League’s Bengaluru season, commentated the final of that AI tournament back in August 2025. He rated Grok 4’s play at around 800 Elo, comparable to a beginner who had only just learned the rules. He described its level as “a talented kid who doesn’t know how the pieces move.” He rated o3 somewhat higher, around 1200, still a modest hobbyist level, well short of a serious club player. Neither model came close to solving anything. The most advanced chatbot available in 2025 was, by a world champion’s own account, closer to a beginner than a threat to the game itself.
Chess Bharat’s Honest Take on This!
Nobody actually knows whether chess will ever be solved. It may happen someday, if computing power keeps growing in ways that are genuinely hard to picture from here. It may also stay permanently out of reach. The gap between today’s best engines and a full mathematical proof is not small, and closing it may take far longer than 10 years, if it happens at all. Both outcomes remain honestly possible, and nobody, not Musk, not Chess.com, not Kasparov, can say for certain which one wins.
What is certain is that the answer to that question changes very little about how people actually feel about chess. People still play checkers for fun despite it being fully solved. People still play chess despite engines like Stockfish already outplaying every human on the planet by a wide margin. Whether or not a definitive answer ever arrives, the game itself has already survived every computer that has tried to end the conversation, and it will likely keep doing so.
Quick Answers
1. Has chess been solved?
Not entirely. Every chess endgame with seven or fewer pieces on the board has already been fully solved through computer tablebases, complete since 2018. The full game, starting from the opening position with all 32 pieces, has never been solved, unlike checkers, which was weakly solved in 2007.
2. Why did Elon Musk and Chess.com argue about chess?
Musk predicted chess would eventually be fully solved by computers, and Chess.com’s X account replied “skill issue,” starting a public exchange about chess’s mathematical complexity.
3. Is chess more complex than the number of atoms in the universe?
The total number of possible chess games, around 10^120, does exceed the estimated 10^80 atoms in the observable universe. The number of legal chess positions, around 4.6 x 10^44, does not.
4. What is Elon Musk’s actual chess rating?
There is no verified answer, since Musk has no publicly recorded rated games on any major chess platform. On Chess.com’s own community forums, some users have informally guessed his strength at under 1000 Elo, a beginner level, based on his public comments about the game rather than any actual match data. That figure is a casual community guess, not a measured rating, and should be treated as opinion, not fact.
