What happened this week in AI by Louie The ongoing race between open and closed-source AI has been a key theme of debate for some time, as has the increasing concentration of AI research and investment into transformer-based models such as LLMs. As we have discussed, there have been some signs of open-source AI (and AI startups) struggling to compete with the largest LLMs at closed-source AI companies. This is driven by the need to eventually monetize to fund the increasingly huge LLM training costs. This is only clearer with this week’s news of Microsoft and OpenAI planning a >$100bn 5 GW AI data center for 2028. This would be its 5th generation AI training cluster. They are currently part way through Gen 3 deployment, while Gen 4 is due in 2026.
This AI newsletter is all you need #93
This AI newsletter is all you need #93
This AI newsletter is all you need #93
What happened this week in AI by Louie The ongoing race between open and closed-source AI has been a key theme of debate for some time, as has the increasing concentration of AI research and investment into transformer-based models such as LLMs. As we have discussed, there have been some signs of open-source AI (and AI startups) struggling to compete with the largest LLMs at closed-source AI companies. This is driven by the need to eventually monetize to fund the increasingly huge LLM training costs. This is only clearer with this week’s news of Microsoft and OpenAI planning a >$100bn 5 GW AI data center for 2028. This would be its 5th generation AI training cluster. They are currently part way through Gen 3 deployment, while Gen 4 is due in 2026.