The component crisis continues to wreak havoc across the computer industry, and its latest victim is Intel's top-tier workstation GPU, the Arc Pro B70. The Battlemage card launched with an MSRP of $949 earlier this year and has been regularly available for $1,000... up until now. The Arc Pro B70 has been hit with a 30% price increase in the States, a 26% increase in Europe, and a whopping 48% increase in Korea this month.
In the US, the ASRock Creator variant of the card is currently listed on Newegg and Amazon for $1,299. Price trackers confirm it was available for $999 just last month and received the hike at the beginning of August. Going from $999 to $1,299 constitutes a 30% increase practically overnight. Another retailer, B&H, actually has the Arc Pro B70 listed for an insane $1,779, but that's the reference design from Intel.
Beyond North America, Europe has seen a similar hike despite the GPU already being overpriced there. In Germany, the Arc Pro B70 has been available for roughly 1,200 Euros, or almost $1,400, across different vendors since June. Geizhals' price tracking now shows the average price of the card at around 1,500 Euros instead, which works out to $1,735. That's a 26% increase between the two recorded prices.
Moving to South Korea, the price hike is the most egregious with a 46% jump from late July to today — less than one month. Danawa, a price aggregator in the region shows the Arc Pro B70 cost 1,889,980 KRW, or $1,334, on 21 July. It rose from that to 2,798,980 KRW, or $1,975, by 16 August. This was less of a sudden hike since the graph below shows an incremental curve over the past few weeks, but it's still absurd. Intel's official pricing is actually listed even higher, at 2,817,000 KRW ($1,988), before discounts.
The Arc Pro B70 comes with 32GB of GDDR6 ECC VRAM, making it extremely desirable for AI workloads, in particular, inference. It also happens to be Intel's top-end Battlemage GPU since the company did not launch any gaming-focused flagship this generation. Any card performant in machine-learning tasks is worth its weight in gold today, as proven by constant, silent price hikes that don't seem to be stopping anytime soon.