We got to Intel’s analyst event late – a long drive with some serious traffic disruptions. But we were getting regular updates from friends with better commutes. Several suggested we just turn around – it was clear Intel was not announcing much. The Street seemed to share this sentiment with Intel stock down more than the market which itself was not having a great day. People were expecting more, so in this case no news was bad news. True, they gave us a peak at their product roadmap. And they unveiled some interesting packaging technology. But that was about it. Almost everyone we spoke with agreed that the only real ‘news’ was a confirmation that Intel’s manufacturing processes were on track. Good to hear, but not new,
By this point, we had arrived and the sunk cost fallacy took hold. We stayed and did not uncover anything stunning, but we still learned some interesting things. We see this event out as a preview for how Intel is going to position its next generation of products.
Intel faces three challenges: 1) Fix their manufacturing process; 2) Compete with products built on those new processes; 3) Build Intel Foundry Services (IFS). Challenge #1 seems to have been de-risked, assuming they keep everything moving on track. We did not got much discussion of #3, this was not the venue for that, but Challenge #2 was very much the focus.
For years, Intel has been able to coast on its lead in manufacturing processes, their chips were just going to be better because they were further along the Moore’s Law curve. When that stalled five-ish years ago, they began to lose share and have been running on phones for a while now. As they come back up to speed, they are going to need to find new ways to compete, even if they can radically close the gap with TSMC. The world has just moved on while Intel was stuck in amber. So finding a new position for its product is critical to their future success. What is going to differentiate Intel in the future?
Intel has to fight this battle on multiple fronts. They need to clarify their position in laptops and PCs. They have to shore up their market share in data center CPUs against AMD. And they need a strong AI story.
For PCs (aka client), Intel is betting on a future in which all CPUs just come with AI accelerators built-in . This is the direction that Apple has already moved with its M-Series processors, and both AMD and Qualcomm have talked about this extensively. Client is the one market where Intel has been able to hold onto share the best, end-users are not as performance sensitive as data center customers. The Intel brand and its channel relationships still carry a lot of weight here. But Intel’s positioning is telling – they think that consumers will not even notice that AI is included in their devices – it is just one more thing that makes their laptops run. This makes sense logically, but viewed from a different angle, it seems to imply that Intel does not think it will have much of a performance advantage in the space. If they benchmarked better than the competition, they would be talking that up heavily. We also found it very interesting that when they spoke about competition they called out both AMD and Qualcomm. Appearing in competitors’ slides is powerful validation. And while we think Qualcomm faces a lot of challenges cracking open the PC market, Intel has now convinced us to take them much more seriously.
For the data center, Intel is taking two approaches. First, they are pushing hard on their installed base of software partners. Intel’s 20+-year hold on the data center gave them a formidable competitive barrier in the form of all that 3rd party software optimized to run on x86 CPUs. They went out of their way to point this out. One presenter conceded that Intel was several years behind the competition but already had many more times the software coverage than those competitors. This is a strong point in their favor, albeit one that has seriously eroded in recent years. They also started to talk about Total Cost of Ownership (TCO). Past Intel marketing focused on performance – their chips just ran faster. But data center customers care about a lot more than just performance, they also have to factor in power which is a much tighter constraint on data centers than chip capex. Intel pivoting to a TCO message is telling in that it reflects the realities of the market which dictates a change in their approach, even if it means adopting TCO language used by the many new entrants to the market.
And then there is AI. Here our sense is that Intel is going to talk a lot about freedom, choice and even ‘democracy’. These are all thinly veiled attacks on Nvidia whose CUDA software is seen as a powerful lock-in for their GPUs. Intel has this new slogan “Siliconomy” which we struggle to define, but is attached to a lot of this verbiage. We think the thrust of their fight against Nvidia will focus on the flexibility of Intel’s offerings which do not require lock-in to other Intel parts in contrast to Nvidia which is heavily favoring customers who buy their CPUs, GPUs, networking and full-featured offerings, to say nothing of their nascent software ambitions. This is smart, but it is also shaky ground. Nvidia is notoriously flexible and agile, their lock-in is not reliant on CUDA and their success of late has as much to do with the performance of their products, something which still challenges Intel.
To be clear, Intel is still in early days in sharpening its messaging. Much of this will change as their products shape up. This was evident in the lack of unified messaging with different Intel representatives voicing slightly different messages. For instance, on the topic of AI Training, every Intel presenter hastened to distance themselves from the market or concede that was not an area of focus. This is sensible given Nvidia’s bastion, except in the middle of the show floor Intel was running a Training demo of their new Gaudi 2 GPUs in direct comparison to Nvidia H100. Which one is it?
That all being said, we reiterate that the biggest challenge facing Intel is its own culture. To what degree has the organization recognized the new realities of its markets? We know that senior management understands the depth of the problem, but it is not clear how far down that wisdom has penetrated. For instance, in one session, an audience member asked the presenter how their product compared to Apple’s M-Series CPU. There are many viable answers to this difficult question, but for some reason the presenter gave the worst one possible. He claimed that “We do not compete with Apple.” PCs and Macs are just two different markets with different customers. I imagine Intel’s customers do not feel that way. Nor do the tens (hundreds?) of thousands of end-users who switched from Windows to Mac last year, a fact which Apple points out on their earnings call. Aside from being factually incorrect, this statement raised our hackles as it harkens back to the old days of Intel, with blinders on blocking out the rest of the world.
All in all, Intel seems to be in a much better position than just a year ago. They have a new level of confidence, some of which they have earned. We just hope they do not have too much confidence just yet.