<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>devtake.dev — MTIA</title><description>Articles on devtake.dev covering MTIA.</description><link>https://devtake.dev/</link><language>en-us</language><item><title>Hot Chips 2026: five bets on serving agents cheaply</title><link>https://devtake.dev/article/hot-chips-2026-agentic-silicon/</link><guid isPermaLink="true">https://devtake.dev/article/hot-chips-2026-agentic-silicon/</guid><description>Intel, IBM, Meta and SiFive each showed a different answer to serving model calls cheaply. Two are already in production, one is a 2027 Xeon.</description><pubDate>Tue, 25 Aug 2026 12:45:00 GMT</pubDate><category>hardware</category><category>ai-chips</category><category>ai-infrastructure</category><category>semiconductor</category><category>intel</category><category>ibm</category><category>custom-silicon</category><category>risc-v</category><category>hardware</category><author>hiro-tanaka</author></item><item><title>Meta just committed 1 gigawatt of custom AI chips with Broadcom through 2029</title><link>https://devtake.dev/article/meta-broadcom-custom-ai-chip-deal/</link><guid isPermaLink="true">https://devtake.dev/article/meta-broadcom-custom-ai-chip-deal/</guid><description>Meta and Broadcom expanded their partnership to co-develop four generations of MTIA chips on a 2nm process. Here&apos;s what the deal includes and why Nvidia should care.</description><pubDate>Thu, 16 Apr 2026 08:30:00 GMT</pubDate><category>hardware</category><category>meta</category><category>broadcom</category><category>ai-chips</category><category>mtia</category><category>custom-silicon</category><category>nvidia</category><category>infrastructure</category><author>hiro-tanaka</author></item></channel></rss>