Micron’s new partnership with Anthropic comes as investors question the pace of AI spending and hardware demand following a selloff across parts of the AI sector. The agreement combines technology collaboration, supply commitments and a strategic investment as companies race to secure critical infrastructure for advanced AI systems.
BNN Bloomberg spoke with Matt Bryson, managing director of semiconductor and hardware research at Wedbush Securities, about the significance of the Micron-Anthropic partnership, AI infrastructure demand, memory supply constraints and the competitive landscape among leading AI developers.
Key Takeaways
- Bryson said demand for AI infrastructure remains robust despite recent weakness in AI-related stocks and concerns about spending levels.
- Memory remains one of the tightest parts of the AI hardware supply chain, making supply agreements strategically important for AI developers.
- Anthropic’s partnership with Micron helps secure access to critical infrastructure needed to support future AI growth and data centre expansion.
- Rising AI token costs may affect adoption decisions in the near term, but Bryson expects technology improvements to lower costs over time.
- Competition among major AI developers remains intense, with access to hardware and supply chains becoming an increasingly important advantage.
Read the full transcript below:
ROGER: Some shares are diving right now, including Micron, amid a broader selloff for AI-related firms triggered by cost concerns and a slower pace of production. Micron is now up seven per cent. This comes one day after Micron announced a partnership with Anthropic for the adoption of Claude throughout its operations. Here to talk about that is Matt Bryson, managing director in semiconductor and hardware research at Wedbush Securities. Matt, thanks very much for joining us on an interesting day, to say the least.
MATT: Yeah, the selloff is kind of crazy around AI concerns. If you step back a couple of weeks, it certainly feels — and the conversations I’ve had suggest — that AI demand, if anything, is stronger than ever.
ROGER: What’s driving this? Is it concern over what we saw in that report out of China about SK Hynix, or is this broader?
MATT: I’m not sure. I would say that when you get concerns around hardware spending, memory tends to act a bit worse because you have this supply-demand dynamic. If hardware demand falls off for memory, you have to worry that you have more supply than demand, and prices are going to fall. That has significant ramifications for memory stocks beyond what you see with other hardware names. So I think that’s partly why you’re seeing memory react worse, but I’m not quite sure what’s driving the concerns today versus yesterday.
ROGER: Michael, we’ll let you jump in here.
MICHAEL: Hey, Matt, how are you doing? What’s the gain for Micron partnering with Anthropic over other AI companies?
MATT: I don’t know that there’s a gain in partnering with Anthropic per se. I would say that if you look at what Anthropic is doing — for instance, working with Google around TPUs — buying TPUs is one thing, but you need a whole lot of other hardware infrastructure to support data centre buildouts. The single tightest point in the hardware supply chain right now is memory. From an Anthropic standpoint, you can certainly see why it would want to have a closer relationship with Micron. Part of what they’re doing is looking at diversifying their data centre portfolio, if you will.
ROGER: Is this triggering concerns about more circular buying and circular investing?
MATT: I think we’re certainly seeing a lot of that. Having said that, if you looked at something like token prices through May, token pricing up until June was increasing, which tells you that regardless of whether investments are being financed in a circular manner, there’s simply more demand for AI resources.
I would say the same thing about hardware. My supply-chain checks from late May through mid-June were all showing surprisingly robust demand for Nvidia GPUs in particular. Remember, we’re at the end of the hardware cycle for them. Typically, at this point, you’d be starting to see people looking more toward sourcing Vera Rubin as opposed to Grace Blackwell, and that just doesn’t seem to be the case.
Rather, for the first time in a long time, I’ve run into Nvidia simply not having enough supply.
ROGER: I want to jump in quickly on token pricing. We’re starting to hear more and more about it. Up until now, everything has been AI, AI, AI, but now we’re actually hearing about a price tag. Is that going to be a factor? Will it make companies take a second look and maybe not move as fast with AI?
MATT: In the sense that it’s becoming more expensive from a point-of-inference perspective, I definitely think that’s something companies need to think about. What is the return on investment on this spending? It’s a little hard to tell because I think a lot of companies are at the very beginning of their AI journey and don’t yet know exactly what the returns are.
Having said that, over time the expectation should be that token prices fall. Equipment gets more powerful, so the cost of producing a token should move lower. As such, whatever you’re paying now, you should be paying less sometime in the future.
I think it’s a concern for adoption at any single point in time, but from a longer-term perspective, AI is always going to get cheaper. It shouldn’t be viewed as a long-term hindrance to adoption.
ROGER: Michael?
MICHAEL: Matt, I had a quick question. Is the end game for the three big large language models — let’s say OpenAI, Google and Anthropic — does this deal give Anthropic any sort of edge over OpenAI or Google?
MATT: It’s really hard to say. If you remember, OpenAI had some commentary late last year around discussions and long-term agreements with Korean memory suppliers. I think those were more future-looking and didn’t necessarily have set terms.
It is important to develop and standardize your supply chain, particularly today, in terms of ensuring you have access to the components you need. When you look at technology markets, we tend to see one or two players emerge as leaders — Google in search, Microsoft and Google in desktop applications, Android and iOS in smartphone operating systems. So it’s really important to get ahead now.
Having said all of that, I would say the best-positioned company right now is Nvidia. Nvidia is sourcing memory and selling solutions that already include memory. To some extent, if you think about OpenAI leveraging predominantly Nvidia hardware, it’s benefiting from Nvidia’s sourcing. Anthropic can be doing the same, and Google can be doing the same.
So, is it important? Yes. Is it something that potentially weighs on the other players? It depends. There are other ways to get around the memory shortage, if you will, right now.
ROGER: All right, we’ve got to wrap it up there. Matt, thanks, as always, for joining us.
MATT: Thank you.
ROGER: Matt Bryson, managing director of semiconductor and hardware research at Wedbush Securities.
---
This BNN Bloomberg summary and transcript of the June 23, 2026 interview with Matt Bryson are published with the assistance of AI. Original research, interview questions and added context was created by BNN Bloomberg journalists. An editor also reviewed this material before it was published to ensure its accuracy and adherence with BNN Bloomberg editorial policies and standards.
