GTA 6 has 6.45 million pre-orders — except nobody at Rockstar counted them

Grand Theft Auto VI has 6.45 million pre-orders and has already taken $630 million. That is the sentence running across the games press this week, and the part it leaves out is who produced the number. It was not Rockstar. It was not Take-Two. It is a reading from a live tracker run by Sensor Tower, a mobile and games analytics firm, and it is an estimate built from a model.

That does not make it wrong. This site spent yesterday taking apart a figure whose inputs did not multiply into its output, and it is worth saying plainly that this one is not that. Sensor Tower's figures are internally consistent to the dollar — we checked, and the working is below. The uncertainty here is not in the arithmetic. It is in the measurement underneath it, and in the three words "digital pre-orders" that most headlines drop.

Where the number comes from

Sensor Tower publishes a public Grand Theft Auto VI sales tracker that updates as pre-orders accumulate. The figures quoted this week are readings taken from it, relayed by outlets rather than announced by anyone.

The platform split published alongside the 20 October reading:

Rockstar and Take-Two have confirmed none of this. No official pre-order number for Grand Theft Auto VI has ever been published, and as set out below, that is a deliberate decision rather than an oversight.

The arithmetic does hold

Most circulating estimates fall apart when you multiply them out. This one does not, and the reason is worth following, because it tells you something the headline does not.

Sensor Tower's analyst Karl Kontus reported in August that 89% of tracked pre-orders were for the $99.99 Ultimate Edition rather than the $79.99 Standard. Price that mix:

That is the published $630 million, to within rounding. The platform rows reconcile the same way: 4.80 million at that average is $469 million against $468 million reported, and 1.66 million is $162 million exactly. The 13 October reading divides out at $97.70, the same model a week earlier.

So the figures are one coherent set produced by one method, not three numbers assembled from different places. What follows from that is less comforting than it sounds: every row above depends on the same estimate of how many units were sold, so none of them independently confirms any other. Consistency is evidence of a single model applied carefully. It is not evidence that the model is right.

What it does not count

Three exclusions, and the first is the one that matters most.

Digital only. The tracker measures digital storefront pre-orders. Physical pre-orders are not in it. Normally that would be a large hole — and here it is unusually small, because Grand Theft Auto VI ships with no disc at all. The boxed product is a card with a download code on it, and Take-Two's business was already described by Strauss Zelnick in August as "well over 90%" digitally distributed. The physical channel is real but narrow. It is still missing, and nobody quoting 6.45 million says so.

Some regions may be absent. The reporting repeatedly carries a line that coverage is not global. Which storefronts and territories are in the model has not been published anywhere we could reach.

Returns are disputed. Accounts disagree on whether the totals are net of cancellations — some say the figures are after returns, while GameSpot's write-up of the 6.08 million reading said it apparently still included refunded copies. We could not resolve this, and it matters more than it looks, for the reason in the next section.

The company that has the real number will not give it

This is the part that makes the estimate newsworthy rather than merely interesting. Take-Two knows exactly how many pre-orders exist. It has been asked, directly and repeatedly, and it has said no on the record.

Strauss Zelnick's position, given to Bloomberg and reported by GameSpot and PC Gamer, has three parts:

• Pre-orders "vastly" surpassed the company's internal forecasts — a direction, with no number attached.
• Raw figures could mislead investors, because the company does not know the shape of the demand curve. In PC Gamer's rendering of his reasoning: "It's just never happened before."
• A pre-order is not a sale. "We haven't sold one unit yet. You can cancel a preorder."

He has also warned that the surge may represent a pull forward of purchases that would have happened later anyway, which is an argument that a big pre-order number tells you about timing rather than about total demand.

Earlier in the year he called the level "unprecedented and astonishing" on the August results call, and said the mix was skewing towards the premium edition — the only official statement that points in the same direction as Sensor Tower's 89%, without endorsing the figure.

Read together, that is a company telling you the number exists, that it is very large, and that publishing it would create a false impression. Into that gap a third-party model has supplied a figure, and the figure is being quoted as though the refusal never happened.

Two numbers that are older or softer than they look

The 89% edition split is from August. It was reported when the tracker showed about 4.77 million pre-orders. October coverage repeats "the vast majority" without publishing a fresh split, so the share quoted this week is not a current measurement. Kontus's own platform breakdown at the time was 90% on Xbox and 88.5% on PlayStation, and he noted how far outside the norm it sits: a 50% premium share during pre-orders would be "a great number", where typical premium shares "tend to be at around 10-20%".

The 25 million projection is a forecast, not a reading. Kontus has projected roughly 25 million pre-orders by launch, worth around $2 billion, with most arriving in the final week before release. He also expects the Ultimate share to fall towards an even split as the late, less committed buyers arrive. Both are modelled expectations about behaviour that has not happened yet, and the second one quietly undercuts any attempt to apply today's 89% to tomorrow's volume.

When an official number could actually exist

Not on 9 November. Take-Two publishes second-quarter results that day, and the quarter being reported closed on 30 September — seven weeks before the game goes on sale. Pre-order money is deferred revenue; it does not become net bookings when somebody places an order. The first Take-Two filing that can contain a Grand Theft Auto VI sales figure is the third-quarter release in February 2027.

What 9 November can produce is commentary: another question to Zelnick, and another refusal, or possibly the first qualitative update since August. The company has never disclosed a pre-order total for any title and there is no sign it intends to start.

What is not confirmed

What to do with the figure

Treat 6.45 million as what it is: one firm's estimate of digital pre-orders on a given day, produced by a consistent method, missing the physical channel and an unstated set of regions, and unconfirmed by the only two companies in a position to confirm it.

That is not nothing. A model this internally tidy, tracking a direction the publisher has independently described as unprecedented, is reasonable evidence that pre-orders are enormous. It is not a sales figure, it is not a record, and it is not news that Rockstar announced — and the gap between those sentences is the whole of this story.

If you are deciding which version to buy rather than counting other people's orders, the two editions and the $20 between them is the more useful page, and preload opens on 12 November ahead of launch on 19 November.

A note on sourcing

Rockstar's own site, Sensor Tower's tracker and most of the games press are unreachable from where this was written. We have not read Sensor Tower's tracker or its methodology note first-hand, and the readings above are taken from reporting of them. Zelnick's quotations are as rendered by GameSpot and PC Gamer, which carried his Bloomberg remarks and the August results call. The 89% edition split and the 25 million projection are attributed to Sensor Tower's Karl Kontus in that same reporting. Where accounts differ — on whether returns are netted out — both readings are given above rather than one being chosen.