HPE has direct visibility into memory supply chains in a way that most companies don't. When they say the shortage runs longer than expected, that's not speculation. It's procurement reality. HPE's server segment actually declined 2.7% year over year despite the company being at the center of enterprise AI infrastructure, with memory constraints identified as the reason.
Moore tells that the vast majority of AI projects never make it to production. There is a good reason for that. This reason has to do with data, he says: "If you race ahead and do all the sexy AI stuff without getting the data right, then you're probably going to fail." This mirrors the failed promises of big data and enterprise data warehousing from previous technology cycles. It's a shame we apparently don't learn from those experiences.