AI Newsround – Broadcom & Meta

Broadcom FQ1 26 – Better bargains elsewhere

• Broadcom reported good FQ1 26 results and made some bold predictions as its AI business is growing faster than Nvidia’s, although from a much smaller base.
• FQ1 revenues / Adj-EPS were $19.3bn / $2.05, just ahead of estimates of $19.26bn / $2.03, but it was on the next two years where the focus was placed.
• AI revenues are currently growing at just over 100% YoY, and the expectation is that around $40bn of annualised AI revenues will become $100bn in 2027.
• Like Nvidia, Broadcom currently has more demand than it can handle, and with Google massively increasing its spending on TPUs and Meta (see below) doing more customised accelerators, I think it has pretty good visibility in terms of what fiscal 2026 is going to look like.
• Hence, its revenues are more of a factor of the capacity that it has booked at TSMC, meaning that it knows pretty much where sales are going to land at least 12 months in advance.
• This means that it is carefully managing expectations such that it can just beat them, which will provide support to the valuation of the shares.
• Here, FQ2 26 revenues / EPS are expected to be “approximately $22bn” which is again just ahead of forecasts at $20.8bn.
• The current growth rate means that AI revenues should be around $50bn this year, making $100bn in 2027 not an enormous stretch.
• However, this assumes that the current spending splurge continues unabated and with rapidly rising debt and deteriorating balance sheets, I am getting increasingly concerned with regard to AI’s affordability.
• As a result, I do see some risk to the 2027 numbers, which should be factored into the valuation of the shares.
• Here, the shares are trading on 30.2x FY 2026 and 20.8x FY2026, which is now significantly more expensive than Nvidia, which is on 22.1x FY2026 and 16.8x 2027 and I suspect represents a lower risk option.
• It is also much more expensive than the memory makers Samsung (which I hold), SK Hynix and Micron, meaning that there would have to be something pretty special about Broadcom to make me want to buy it.
• Hence, I see no reason to sell my positions and move into Broadcom, about which I remain pretty ambivalent.

Meta Platforms – Custom free-for-all.

• Meta Platforms is going to spend a vast amount of money on AI data centres this year, a good proportion of which will be on custom silicon, creating a great opportunity for existing players and newcomers.
• Meta Platforms’ CFO Susan Li has stated that it has recently reached deals with a number of chipmakers to help it design and build customised silicon for its AI datacentres.
• Meta Platforms is different from its peers, such as Google, Microsoft and Amazon, in that it does not offer its compute for sale to third parties, meaning that it is only running its own workloads.  
• This means that customised silicon makes more sense for Meta than others to use custom silicon, as software and hardware can be fully architected to work together.
• I suspect that this is why the market has been flush with rumours for some time about other players doing deals with Meta Platforms for custom AI silicon.
• Consequently, I think that there is a good chance that both Arm and Qualcomm announce deals with Meta at some point and I would not be surprised to see Broadcom, MediaTek and a number of others also get a piece of the action.
• With around $135bn of capex up for grabs this year alone, there is plenty of space for everyone to get a meaningful piece of the action.
• I remain increasingly concerned with the deterioration of Meta’s balance sheet, its falling margins and the weakness of its position in AI.
• Hence, the place to look for a positive return on Meta’s AI, is not Meta but all of the companies that look set to benefit from it.
• Here, I own Qualcomm, where this would be icing on the cake, and I am also taking a look at MediaTek, although I appear to be a bit late to the party.
• Arm is also coming into range as its underlying financial performance has outperformed the stock price, meaning that the valuation is much more reasonable these days.
• I would not touch Meta, as I think it could easily underperform from here.