SanDisk and its manufacturing partner Kioxia announced Thursday that they plan to spend more than $31 billion on memory chip factories in Japan by 2032, and Wall Street is doing a double-take. Three weeks earlier, SanDisk's own CEO told investors the company was moving away from large capacity expansions.

Both things are apparently true at once. That tension is worth understanding, because the outcome shapes where the world's flash memory gets made, and who controls it.

What actually happened

The two companies said they would invest the $31 billion in their shared plants in Yokkaichi and Kitakami, the Japanese facilities where they produce NAND flash memory, the type of chip that stores data in your phone, laptop, and camera. The investment also covers related technology development. For context, the two companies have spent just over $50 billion in Japan across more than 25 years of partnership. This new plan would spend roughly 60% of that total again in just six years.

The plan depends on Japanese government support, which means it won't move forward without a subsidy commitment from Tokyo.

Who actually writes the checks

SanDisk is not writing a $31 billion check. The two companies manufacture through a shared structure called Flash Ventures, which runs eight facilities in Japan. SanDisk holds just under a 50% stake in those entities. Under its financing obligations, SanDisk is on the hook for roughly half of whatever capital the joint ventures decide to spend, to the extent that the ventures' own cash flow can't cover it.

Neither company has spelled out its exact share of the $31 billion. But if roughly half of the total flows through Flash Ventures, something close to $1.3 billion per year lands on SanDisk, before any government contribution. That is a real number, even if it is smaller than the headline figure suggests.

The contradiction SanDisk has not fully resolved

Here is what makes this announcement strange. On August 5, three weeks before Thursday's news, SanDisk CEO David Goeckeler told investors: "We grow supply primarily through nodal transitions rather than wafer additions, delivering mid- to high-teens bit growth." His CFO added that capital spending would fall to roughly 6% of revenue in fiscal 2027.

In plain terms: the company said it was growing the amount of memory it can produce by improving manufacturing technology, not by building more physical capacity. The pitch to investors was that this approach keeps spending lean.

A $31 billion factory investment plan sits awkwardly next to that pitch. The two are not necessarily incompatible. Much of the $31 billion is likely earmarked for technology upgrades and infrastructure rather than raw new capacity. But SanDisk has not clearly reconciled the two stories, and investors are entitled to ask which one governs actual decision-making.

Why the location matters

Japan is not a neutral backdrop here. The global chip industry spent the last several years discovering how badly concentrated semiconductor production had become, mostly in Taiwan and South Korea. Governments in the United States, Europe, and Japan have been spending heavily to pull manufacturing closer to home or into allied territory.

Japan's government has been particularly aggressive about supporting domestic chip capacity. The conditional language in Thursday's announcement, that the plan depends on government support, signals that Tokyo's subsidy commitments are likely a live negotiation rather than a formality.

For the broader memory chip market, a $31 billion bet on Japanese production by two of the industry's major players reinforces Japan's position as a counterweight to South Korean and Taiwanese dominance. That matters for supply chain resilience, for geopolitical leverage, and eventually for the price and availability of every device that runs on flash memory, which is most of them.

The capital-light story SanDisk told investors may still be true at the company level. But at the industry level, the memory business is clearly entering a very expensive phase.