{"id":94315,"date":"2026-02-16T12:30:00","date_gmt":"2026-02-16T09:30:00","guid":{"rendered":"https:\/\/forklog.com\/en\/?p=94315"},"modified":"2026-02-16T12:33:19","modified_gmt":"2026-02-16T09:33:19","slug":"the-silicon-curtain","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/the-silicon-curtain\/","title":{"rendered":"The silicon curtain"},"content":{"rendered":"

The age of digital abundance\u2014when any enthusiast could assemble a home server to rival a small firm\u2019s kit\u2014is drawing to a close. Owning cutting-edge hardware is taking on an air of elitism amid rising memory-chip prices and lengthening pre-order queues.<\/p>\n

In a new ForkLog piece we examine why graphics cards have become the lifeblood of the AI industry, why Nvidia seems to love gamers less, and why freelance designers are renting capacity from cloud DC<\/span>s. The larger question: how will the chip crunch affect blockchain decentralisation, where SSD<\/span>s and DRAM<\/span> are often pivotal?<\/p>\n

Techno-feudalism or a passing squall<\/h2>\n

Judging by recent remarks from AI-industry chiefs and memory-chip makers, the age of the powerful personal computer (PC) may be ebbing.<\/p>\n

The tech world is still chewing over a 2024 talk<\/a> by Amazon\u2019s founder Jeff Bezos, who likened the PC to a home generator in the era of grid electricity. Some now hail him as a seer.<\/p>\n

The latest hardware has become the prime compute for training and serving LLM<\/span>s. AI is stripping warehouses of HBM<\/span> chips, whose makers once catered chiefly to consumer SSDs and RAM. As component prices climb, the market could shed an entire class of budget devices as soon as this year.<\/p>\n

In early February TrendForce researchers raised<\/a> their chip-price forecasts, expecting contract prices for consumer DRAM to jump by 90\u201395% in the first quarter of 2026 amid the AI boom. The previous estimate was 55\u201360%.<\/p>\n

Training LLMs also demands vast troves of data. Enterprise buyers have snapped up SSDs of 2 TB and above with high write endurance. Silicon vendors, seeing fatter margins in AI, are reorganising capacity.<\/p>\n

At the end of 2025, Micron Technology\u2014the memory leader and once a staunch defender of desktops\u2014announced<\/a> it would shutter its Crucial consumer line. Production will cease in the second quarter of 2026 after nearly 30 years of the brand\u2019s life.<\/p>\n

Micron also plans to ramp HBM output. The company invested<\/a> $9.6bn in new capacity in Hiroshima, Japan.<\/p>\n

On 12 February Samsung Electronics said<\/a> it had begun shipping advanced HBM4 chips to unnamed customers\u2014an effort to close the gap with rivals in critical components for Nvidia\u2019s AI accelerators, including SK Hynix.<\/p>\n

The world\u2019s largest chipmaker sits in a bind: it is a key memory supplier to Nvidia while also leading in smartphones and consumer electronics. It must preserve rich-margin AI contracts without weakening its gadget franchise.\u00a0<\/p>\n

In September last year Samsung Semiconductor tried to strike a balance. The company confirmed<\/a> that its top-end graphics-memory lines\u2014GDDR7\u2014can serve gamers and content creators as well as professional workstations.<\/p>\n

These chips power Nvidia\u2019s gaming flagship, the GeForce RTX 5090<\/a>. Unveiled in January 2025, the card remains the unchallenged leader; last year\u2019s $1,999 sticker bears little relation to reality. At the time of writing, listings range from $4,000 to $5,000.<\/p>\n

\"image\"
Source: Nvidia.<\/figcaption><\/figure>\n

China\u2019s highly adaptive market, as usual, is seizing the opportunity. According to Nikkei Asia<\/a>, memory makers CXMT and YMTC plan a major capacity build-out.<\/p>\n

In 2027 they aim to open fabs in Shanghai and Wuhan, focusing first on DRAM and NAND<\/span> rather than HBM, as market leaders do.<\/p>\n

Ex-CIO\/CTO of Bitfury Group and Hyperfusion co-founder Alex Petrov argues there is little sense in waiting for prices to fall; better to reallocate costs.<\/p>\n

\n

\u201cDo not wait; we live here and now. If you need hardware for work, mining, a node, it is better to buy now, putting up with high prices, and set aside what you can temporarily do without. Deferred demand by 2028 may be huge and unpredictable; one can only hope for old DDR3\/4 and the arrival of new DDR6,\u201d<\/em> the expert told ForkLog.<\/p>\n<\/blockquote>\n

Why graphics cards?<\/h2>\n

Why did the graphics cards that ran Quake III Arena in 2000 and Fallout 4 in 2015 end up first commandeered by PoW<\/a> mining and then swallowed by the AI industry? The answer lies in the nature of GPUs\u2014best seen in contrast to the CPU.<\/p>\n

A CPU is a genius that can tackle any software task: write poetry, do taxes, run an operating system. But it executes sequentially on each core.<\/p>\n

By contrast, a GPU<\/span> is a factory of thousands of simple workers. Each is less bright than the genius, but they act in parallel.<\/p>\n

To render a frame, you must compute the colour of millions of pixels\u2014millions of identical operations per second. The graphics chip was born for parallelism.<\/p>\n

Something similar happened with PoW mining on GPUs. Mining is a kind of lottery: the device must try billions of nonces per second to find the right hash. GPUs were ideal for this, spurring the first wave of shortages before Ethereum moved to PoS<\/a> in 2022.<\/p>\n

GPUs then proved a boon for AI. Modern LLMs such as ChatGPT or Gemini are, in essence, giant matrices\u2014tables of numbers. Training them is endless matrix multiplication to tune \u201cweights\u201d (connections between neurons).<\/p>\n

The maths that paints water reflections in Cyberpunk 2077 is the same linear algebra that underpins neural-network training. But AI needs not only compute grunt; it needs staggering data throughput. Consumer graphics memory is not enough\u2014HBM has taken its place, pricey and scarce, and now courted by every tech giant.<\/p>\n

Nvidia spotted this in time and, starting with the Volta architecture, began adding \u201ctensor cores\u201d tuned to perform matrix multiplications for AI tasks.<\/p>\n

GPUs by the hour\u2014and the loss of offline<\/h2>\n

For at least the next two years, content producers, video editors, designers, gamers, programmers, AI architects\u2014anyone whose work hinges on powerful hardware\u2014must choose: rent online capacity or pay a steep premium to upgrade a PC.<\/p>\n

With shortages and queues for some components, subscriptions are gaining traction, pushing cloud data centres to become more customer-friendly. A raft of firms now rents flexible compute and GPUs, including Lambda Labs, Vast.ai, Hyperfusion, LeaderGPU, Hostkey and others.<\/p>\n

RunPod offers time on the scarce RTX 5090 at $0.89 per hour.<\/p>\n

\"image\"
Source: Runpod.<\/figcaption><\/figure>\n

Shadow provides a remote desktop without limits on games or pro software for engineers and designers. By contrast, GeForce Now or Xbox Cloud offer less freedom\u2014but at different price points.<\/p>\n

\"image\"
Source: Shadow<\/a>.<\/figcaption><\/figure>\n

Already, with a stable connection, a home smart TV can double as a powerful workstation if you rent the right \u201ciron\u201d. The trade-off: responsibility for quality and uptime shifts from users to owners of tech parks, who may prioritise larger clients and enforce sanctions.<\/p>\n

Petrov noted that data centres guarantee 24\/7 availability, backup power, redundant connectivity and proper maintenance quality.<\/p>\n

\n

\u201cAt the same time, you can quite well keep some things at home or at work. It is just often more expensive and less convenient,\u201d<\/em> he added.<\/p>\n<\/blockquote>\n

According to him, many designers, video editors, producers and artists are already being displaced by AI. At some point they must lean on specialised AI apps that home hardware cannot handle.<\/p>\n

\n

\u201cWhile LLM requirements grow exponentially, you can keep only small models on a phone or at home. Larger expert versions require a different scale, capacity and speeds, which is exactly what cloud data centers provide,\u201d<\/em> Petrov explained.<\/p>\n<\/blockquote>\n

Bitcoin out in front again<\/h2>\n

The entire IT sector depends on components, but for blockchain the chip crunch poses a real threat to decentralisation and to the balance of power.<\/p>\n

\n

\u201cThe rise in memory prices is the consequence of decisions by individual commercial companies. Blockchain nodes are not the only ones affected; prices are rising for all devices with new DDR5 memory: smartphones, PCs, everything. This also forces blockchains to become smarter and more economical, to look for different paths and solutions for this,\u201d<\/em> the Hyperfusion co-founder believes.<\/p>\n<\/blockquote>\n

He highlighted a paradox now dogging PoS networks:<\/p>\n

\n

\u201cProof-of-Stake reduced the energy consumption of mining, but shifted the burden from electricity to memory and disks for businesses and users. In conditions where components have become 3\u20135 times more expensive, PoS chains have found themselves in a \u2018perfect storm\u2019 of reality.\u201d<\/em><\/p>\n<\/blockquote>\n

Blockchains like Ethereum and Solana<\/a> follow a principle of \u201ceasy to create, but extremely costly to verify\u201d. Given many nodes and seven to nine proof steps, the validator barrier to entry in PoS networks is often lower at deployment but higher in operating expenses.<\/p>\n

\"image\"
Technical requirements for running a Solana node. Source: Solana Labs<\/a>.<\/figcaption><\/figure>\n

Petrov notes that in Ethereum each node must keep the entire database of accounts, contracts and balances in near-instant reach. That is tens of millions of constantly updated objects. For speed you need high-performance RAM and NVMe<\/span> SSDs combined in a RAID array<\/span>.<\/p>\n

Nodes must process every block. In high-frequency networks (Solana \u2014 400 ms, Ethereum \u2014 12 s), signature verification and transaction execution demand huge resources. Requirements for full archive nodes are far higher: in Ethereum an archive node calls for 128 GB RAM\/ from 12 TB SSD.<\/p>\n

Falling validator margins as components get pricier creates a fresh centralisation risk. In January the daily number of active nodes on the Solana blockchain fell<\/a> to 800\u2014the lowest since 2021. As participation shrinks, small node operators struggle to cover voting and infrastructure costs unless they have ample delegated stake.<\/p>\n

At the time of writing the network\u2019s Nakamoto coefficient<\/a> fell<\/a> to 19 (in 2023 it was<\/a> 33).<\/p>\n

Within the Ethereum Foundation<\/a>, ideas to lower infrastructure barriers are already under discussion. In May 2025 Vitalik Buterin proposed<\/a> EIP<\/span>-4444, which could materially reduce disk requirements. Nodes would store only the last 36 days of transaction history while retaining the current state and the structure of Merkle trees<\/a>\u2014cutting storage without impairing verification of the live chain.<\/p>\n

In the new \u201csilicon curtain\u201d era, bitcoin remains the \u201cpeople\u2019s blockchain\u201d.<\/p>\n

\n

\u201cIn bitcoin there is no state verification, only <\/em>UTXO<\/em><\/a>, which are easy to cache. The creation phase of PoW mining requires ASIC farms and huge energy-optimised capacity, but validation remains ultra-light. Verifying the result in PoW is extremely simple and fast\u2014that is its beauty. Steps on a validator node: get the block data, check its hash, perform one or two hash operations, compare the target\/difficulty, and everything is clear\u2014yes\/no,\u201d<\/em> Petrov explained.<\/p>\n<\/blockquote>\n

For these reasons, a full bitcoin node can run on a modest server or desktop\u2014and sometimes on newer Raspberry Pis with 4\u20138 GB of RAM. The impact of memory shortages on PoW nodes is minimal. SSD prices are rising, but capacities up to 1 TB remain within reach, he added.<\/p>\n

What next?<\/h2>\n

Petrov argues the era of personal hardware is not over\u2014just different approaches for different jobs:<\/p>\n

\n

\u201cI like the quote \u2018Cloud is someone else\u2019s computer\u2019 \u2014 \u2018The cloud is just someone else\u2019s computer on the network\u2019.\u201d<\/em><\/p>\n<\/blockquote>\n

Industry is scrambling for a way out of the chip crunch, developing new technologies:<\/p>\n