[12:12 Thu,11.December 2025 by Thomas Richter] |
It is common knowledge that AI companies are responsible for the drastic price increase of DRAM
;) Officially, OpenAI intends to use the DRAM for its gigantic Stargate Project, under which a massive sum of 500 billion dollars is to be invested by 2029 to build AI data centers. This involves building data centers across the USA with up to 10 gigawatts of computing capacity. This requires a huge number of GPUs and the associated RAM. Panic in the Memory MarketTo the sudden removal of a significant portion of DRAM production from the free market, another factor contributed to further price increases shortly after the deal: all companies heavily reliant on DRAM, such as OpenAI&s competitors or PC manufacturers, panicked: if such massive deals could be arranged behind their backs – which other secret DRAM deals were still pending? This fear led to – also secretly kept – large-scale purchases of DRAM to avoid being left empty-handed and potentially unable to manufacture their products without the necessary DRAM. This led to a further shortage of memory available on the market. Scarce ReservesAdded to this was the fact that the DRAM market had fewer reserves than usual – the DRAM purchases swept the market clean. The reasons were the Trump tariff chaos, to which companies reacted with less stockpiling, and the steadily falling prices for DRAM – before the OpenAI memory purchase. Furthermore, the major memory manufacturers did not sell their old equipment freed up by the upgrade to the latest technology to smaller manufacturers, as was usual, for fear of penalties from the Trump administration. This equipment was therefore not available for DRAM production. These interconnected factors led to the rapid rise in memory prices starting in October. Raw Wafers Instead of Finished Memory ChipsInterestingly, OpenAI did not buy finished DRAM, but "raw" wafers that still need to be cut and further processed to produce usable memory chips like DDR5 or HBM memory specialized for AI applications. This means OpenAI owns up to 40% of global memory production without being able to use it yet. Moore&s Law Is Dead interprets this as a maneuver by OpenAI to "starve out" the AI competition – without the memory, which lies unused and uncut in warehouses, there are no new GPUs (because a lot of memory is central to AI calculations alongside the pure computing power of the GPU cores), and thus no increase in AI computing power and no larger training capacities (currently the most popular way to further improve the quality of an AI). And when they buy memory, it is many times more expensive than before, costing money that must be raised additionally and unexpectedly. ;) DRAM Wafers behind finished memory chips Eliminating Competition Through ScarcityEspecially since OpenAI has recently come under significant competitive pressure – from China&s DeepSeek and most recently from Google&s new Gemini 3 model – and the enormous investments must pay off, Sam Altman might be trying a different approach to stay ahead of the competition through the memory deal – if not with a better AI, then at least with more computing power. This might work – however, the collateral damage for entire economic sectors and many users is enormous. ;) Daily prices for RAM in a computer store in San Francisco What are the consequences for RAM, SSDs, GPUs,...?Primarily: continued rising prices for memory for the foreseeable future, as there are no stocks and ramping up production through various means takes 6-9 months. Until this production really impacts the market in large quantities, it will be expensive (for example, the price of Crucial 32 GB DDR5 RAM has risen from around 78 Euros in September to  currently 302 Euros – this current memory crisis could potentially last until mid-2027). And the memory giants Samsung and SK hynix do not even want to accelerate the expansion of their production capacities as they fear overcapacity. Logically, all products that use memory chips, such as graphics cards and PCs (at least those from smaller companies that do not have fixed long-term supply contracts or reserves like larger manufacturers), will become scarcer and more expensive. AMD&s GPUs arrive at OEMs without memory and could therefore soon face problems due to scarce memory. But even Nvidia&s upcoming GPUs like the RTX 5080 Super with 24 GB are reportedly postponed to Q3 due to a lack of memory. SSDs will also become more expensive, as their prices follow those of DRAM and are also becoming scarce due to high demand from hyperscalers. Their price increases have already begun – consequently, traditional hard drives will also become more expensive – their price increases have also already started. What to do? Buy quickly?Anyone planning to buy a new laptop, PC, smartphone, or even just memory modules or SSDs soon should probably act quickly, as prices are likely to continue to rise for the time being. Major PC manufacturers such as Dell, Lenovo, and HP plan to increase their prices due to the rising costs of individual components like RAM, GPUs, SSDs, and HDDs, for which reserves are limited. Dell plans to increase many prices by 15-20% as early as mid-December. Chinese smartphone manufacturers Xiaomi and Redmi have already announced price increases of 20% to 30% for customers next year due to high memory prices. And even Samsung&s mobile division is not receiving enough memory as planned for new Galaxy smartphones from its own memory division because other customers like hyperscalers are willing to pay more for it.  more infos at bei www.mooreslawisdead.comdeutsche Version dieser Seite: Ist Sam Altman schuld an der großen Speicherkrise? |
 |
|
|