Broadcom & Snowflake– No Slowdown Yet

AI still driving the numbers

Broadcom FQ3 26 – Fairly well anticipated.

• Broadcom provided further support that the AI story remains intact, but some of the longer-term forecasts are now so high that the capex to support them looks unaffordable.
• This leaves Nvidia as the most credible forecast for 2027, as its forecast depends on a large widening of the customer base as opposed to hyperscalers spending even more money.
• FQ3 26 revenues / adj-EPS were $29.6bn / $3.32, broadly in line with estimates of $29.2bn / $3.22, meaning that all eyes were on the guidance going forward.
• Here, Broadcom put up some pretty big numbers with $115bn in AI chip revenues expected in 2027 and doubling again to $230bn in 2028.
• This is due to large models, a focus on inference and agentic workloads increasing the requirement for CPU’s in the data centre more quickly than GPUs.
• Furthermore, everyone who is operating at scale has either launched their own silicon chips or is in the process of doing so, which plays directly into a sweet spot for Broadcom.
• This increase in estimates leads to a $30 per share EPS forecast (10% above consensus) for 2028, putting the shares on 12.2x FY 2028 PER, which is very reasonable given the growth rate and the expectations for 2029 and 2030.
• However, the short term was a little less exciting with $34.8bn in revenues expected in FQ2 26, broadly in line with expectations.
• This combined with the increasing concerns about how the technology industry can afford $1tn+ in capex is what has kept the reaction to these results under wraps.
• Broadcom could become interesting if the multiple continues to unwind, but at the moment, I am focusing on having a look at Nvidia.

Snowflake FQ3 26 – SaaSpocalypse dying

• Another excellent performance from Snowflake in FQ2 27 hammers another nail into the coffin of the SaaSpocalypse theme that had taken the sector down by 50%, but those that have embraced AI are now beginning to separate themselves from those that have not.
• FQ2 27 revenues / adj-EPS were $1.55bn (up 35% YoY) / $0.62, nicely ahead of expectations of $1.48bn / $0.45.
• FQ2 GAAP EPS was a LOSS($0.55) almost entirely due to stock compensation.
• This is accounted for in the 4% increase YoY in share count, meaning that adj-EPS of $0.62 is a reasonably accurate reflection of the underlying business.
• Snowflake moved relatively early to put AI into its product suite, and this is now showing results, with 9,100 customers now using its coding assistant CoCo, with 2,000 coming on during the quarter.
• This gave the market the confidence to set Snowflake apart from the general software malaise as it has seen an acceleration in growth and the shares have made a new all-time high (up20%) in after-hours trading.
• This is good news for the AI-adopters like CRM, SAP, ServiceNow and Adobe, as it is a further proof point that enterprises will still need software and most companies are turning to their existing vendors to help them with the transition to using AI in a safe and controlled manner.
• The problem with Snowflake is that it remains ludicrously expensive as it is trading on over 150x FY27 PER and 137x FY28 earnings meaning that there are much better value options elsewhere.
• Here I own ServiceNow and Adobe, which between them are 26% of the way to my targets, while Nvidia, which is growing faster than all of them, may yet be cheaper still.
• There is no reason to own this when there is so much else on offer.