Tech Thoughts Newsletter – 23 June 2023.
Market: A more mixed picture this week, with Jay Powell speaking at Capitol Hill and the higher than anticipated BOE rate rise putting rates and inflation back into focus. Quiet results-wise in tech but Fedex – a good off-quarter-reporter and macro bellwether – provided a still uncertain picture on the economy. In general a profit taking week in equities.
Portfolio: we have not made any major changes to the portfolio this week.
Lots of Intel news this week. We don’t own Intel (it’s a structural share loser – which we detail below), but it’s an important company to follow given (1) it’s in the top 3 semiconductor equipment spenders – a sector we’re exposed to; and (2) as a competitor to AMD, Nvidia and TSMC (all owned).
On Wednesday, Intel hosted its Foundry model webinar, detailing their new IDM model “2.0”. Intel’s origin is as an IDM – an integrated device manufacturer – meaning it designs and manufactures its own products.
That’s different from how most of the industry has evolved over the past several decades – most semiconductor companies we talk about are fabless – Nvidia, AMD, Qualcomm, Mediatek, Apple all design their chips but have them made at a foundry (mostly TSMC).
That whole idea of foundry was started with TSMC and Morris Chang who promised “Integrity, commitment, innovation and customer trust” – TSMC would focus exclusively on manufacturing and not create any of its own designs, which was hugely appealing to fabless companies who had before had to rely on taking spare capacity from their competitors and all the risk that entails (having capacity switched off, or even having designs stolen).
Intel however, stuck with its own integrated approach (they arguably suffered from incumbents dilemma) which delivered high margins for a significant period but ultimately got disrupted both from a design perspective (x86 vs ARM on mobile) and from a manufacturing perspective (struggles getting to 10nm which meant TSMC leapt ahead of them).
CEO Pat Gelsinger first set out the Intel IDM 2.0 vision about a month into his tenure, back in March 2021. There were two main features to this: (1) increasing its relationships with foundries to outsource production of its own chips which they’ve done (today, 20-25% of Intel’s silicon is manufactured external – and its new Meteor Lake launching in H2 will be partly on TSMC 7nm); and (2) that it would have a renewed push to become a foundry business – that is to say it would manufacture chips for other semis companies. And this is what the webinar this week was focused more on – in particular around the internal accounting of separating the foundry business.
The changes outlined measures like business units taking market pricing rather than allocated cost, but also that business units would no longer have the luxury to expedite, freely sample and test wafers. That makes good business sense but we’d argue Intel struggled enough with these advantages (it estimates it expedites material tests 2-3 times more than their industry peers..) – it makes their lagging performance over the past decade even more damning..
All of this is ultimately aimed at driving cost down for Intel, which it needs to do – given cash flow and its balance sheet. However, whether this at all helps it to make any headway into the foundry business remains to be seen. We don’t think that changing internal accounting policies changes how appealing Intel is as a foundry partner. Fundamentally it will still be part of Intel and so there will always be a “competition with customer” element, and its history of industry dominance culturally doesn’t help. Most importantly though it is still significantly behind TSMC from a technology standpoint, so for customers like Nvidia and AMD who are in the business of leading from a technology perspective it will be very hard to be persuaded. To be clear on the current leadership: Intel is currently still struggling ramping its 7nm process, while TSMC is shifting its Apple volumes to its 3nm process this summer.
Importantly for us is Intel’s renewed commitment to manufacturing leadership (which is do or die if it wants to try to build a successful standalone foundry business). Its statement on the call that the 2022-2024 net capex intensity will be in the 30s (% of revenue) was positive. The “net” is very important here – we think that means, both ex government subsidies and ex Brookfield’s JV ($15bn announced last summer for their Arizona fab) – that means for example the $33bn Germany fabs (of which the German government will pay $11bn – see below) and the $30bn Arizona fab cost (of which Brookfield will pay $15bn) is net capex number of $38bn but a gross capex number of $64bn. It’s important for our semi cap equipment names – of course it doesn’t matter to them if their revenue and orders are coming from an Intel pot of money or a government pot of money or an Infrastructure partner pot of money – it Is all filling up order books.
Intel also made it very clear how much the manufacturing roadmap is reliant on ASML’s (owned) EUV tools. Gelsinger has been very open about his view that Intel falling behind at 7nm was around it deciding to limit the use of EUV in its manufacturing process (which increased the process complexity hugely around multiple patterning). And again on the call it was made clear that Intel’s return to leading edge manufacturing is totally dependent on it building up leading edge fabs and adopting EUV/High NA – which are already in ASML’s order books. It has (we think) still only 1 EUV machine in production – which it shipped from the US – and maybe 1 more test/R&D tool. For context, we think (as a rough estimate) TSMC has about 80 tools in production. That’s an extraordinary level of catch up required – and we think Intel will be a significant driver of semicap orders for the next 2-3 years – see also below
The other big news from Intel this week was around its various geographic expansion projects and the subsidies it’s receiving:
There was a surprise Sunday night rumour that Intel is going to invest $25bn building a new fab in Israel. It’s a bit of a head-scratcher to us – they will get a grant representing 13% of the total amount which isn’t huge in the context of the other subsidies they’re receiving elsewhere (like in Germany – see below). At the same time their tax rate is going up, not down.
The Germany €30bn investment we mentioned in the letter last week was confirmed in a press release from Intel. They didn’t state explicitly but press reported Germany agreed to the €10bn subsidy (which is up from a prior €7bn). That will be two fabs, the first of which will enter production in 4-5 years and which will now be Intel’s 18A/20A chips. They also announced a $4.6bn investment in a new Assembly and Test Facility in Poland.
As above, it’s all good news for semicap order books. We own ASML, KLA, Applied Materials and LAM Research.
India joining in the semis party.. Modi’s US visit
- Micron (not owned) announced that they will build a new chip testing and assembly plant in Gujarat India. The construction of the new DRAM and NAND plant is expected to start later this year and become operational in late 2024 as part of the first phase of the project. The second phase, which is expected to include construction of a facility similar in size to the first phase, is expected to happen towards the second half of the decade.
- Micron’s investment will be up to $825m, and it will receive total government support – wait for it – up to $2.75bn – from both the Indian central government for 50%, with incentives from the state of Gujarat for 20% of the total project cost.
- It’s a huge number – and India is definitely pushing hard with incentives in order to incentivise companies to invest in their semiconductor production in the country.
Portfolio View: good news for semicap (again…)
EV Batteries also a beneficiary of subsidies
- It’s not just semis getting subsidies – this week it was announced the US issue a $9.2bn loan to Ford for the construction of 3 EV battery factories. It’s the biggest government backing for a US auto maker since the bailouts in the 2009 financial crisis.
- Not only that, cars and SUVs made with domestic batteries will be eligible for the US inflation reduction Act funding, and the US will subsidise the manufacturing of batteries, and potentially offer tax rebates of up to $7,500 per vehicle
Portfolio view: it’s clear that governments across the world have made the decision to incentivise both consumers and auto OEMs to move to EV – we saw that earlier this month with subsidy extensions in China (this week the tax exemptions extension timeline was clarified to last to 2026/2027), and the various government EV production targets. We think it creates a highly competitive environment for auto OEMs but for us what’s interesting in tech is the power semis companies that benefit from a 2-3x semi content increase in the move from ICE to EV. We own NXP and Infineon where we think their power semis remain largely sold out for the rest of the year, and which we think are still benefiting from positive pricing.
Amazon announced its Generative AI Innovation Centre – “3 steps in to a 10km race”
- Amazon (owned) is investing $100m to help companies use generative AI– not a huge amount by any stretch but it’s clearly a tiny piece of what Amazon is doing in totality in terms of its broader AI/AWS/Cloud investments.
- We commented at Q1 results that it looked like AWS might have lost traction vs Microsoft Azure/Open AI, having been the default choice for start ups over the past decade or so – this is part of its attempt to stop that shift from happening.
- AWS CEO Adam Selipsky gave an interview with CNBC which gave some helpful points on the infrastructure investment “GPUs used very pervasively… many many customers come to AWS to use Nvidia based GPUs”.
- When asked “does Amazon have enough Nvidia chips?” his answer was “the entire world wants more chips to do AI.. short term everyone is looking under every rock…” – so again speaking to our view that Nvidia H100 chips are currently supply-gated.
- On capex: “generative AI is certainly an area where capex will increase significantly”.
- Amazon are also developing their own ARM-based AI chips –we’ve commented before that if it makes sense for anyone to develop their own AI chips it’s the hyperscalers, given they have enough utilisation, a big enough user installed base, and specific use cases
Portfolio view: the AI arms race continues – we expect this to be a feature of the upcoming hyperscaler results and for capex (or certainly the capex apportioned to AI) to continue to be revised higher – that will benefit our chip exposure (we own Nvidia, AMD and TSMC).
Consultants/Systems integrators seeing late cycle weakness, future hopes on AI
- Accenture (not owned) reported a top and bottom line beat, though bookings slowed to +4% (from +17%) and book to bills were down to 1.0x/1.1x in consulting and outsourcing.
- Overall top line guidance for the full year was tweaked down while operating margins are expected to come in in line with the prior range. There was weakness in smaller consulting deals and the comms, media and high tech segments. A bit of late-cycle softness for sure.
- Generative AI was of course a key theme on the earnings call and for the systems integrators/consultants it’s about trying to position themselves for the tailwind – as per the announced $3bn investment from Accenture last week.
“I want to turn to generative AI and AI more broadly. No previous technology wave has captured the intention of leaders and the general public as fast as gen AI. We are now embarking on the age of AI, and companies will need to reinvent how they operate with AI at the core. And it is also early. Think of it as the cloud over a decade ago. In fact, in a survey of global executives that we completed just last week, 97% of executives said gen AI will be transformative to their company and industry and 67% of organizations are planning to increase their level of spending in technology are prioritising investments in data and AI.”
Portfolio view: it’s reasonable to think that there will be lots of AI projects which need to be implemented in Accenture’s very big roster of very big enterprise clients, and that them as scale (with an ability to invest $3bn and hire AI talent) player should benefit. At the same time, like all consultants and systems integrators, they have a large “time and materials” type component to their revenue base which AI should arguably put under pressure.
AMD gaining within Oracle’s solution set
- AMD (owned) has been successfully penetrating Oracle’s Exadata solution since 4Q22 and the X9M cloud solution (previously Intel). The release from Oracle around Exadata X10M this week indicates AMD has been chosen as the main supplier or in the best case single supplier for the company in both cloud and on premises solutions.
Portfolio view: We believe this to be a major win for AMD away from Intel and will help AMD sales in 2H23.
AI networking investment alongside strong GPU builds
- Cisco (owned) announced two new AI-focused networking chips, the G200 and G202, and they are currently being tested by customers. G200 and G202 are built to optimise real-world performance of AI/ML workloads – while simultaneously driving down the cost, power, and latency of the network.
Portfolio view: networking infrastructure is a slightly forgotten beneficiary of AI and strong GPU builds – the bandwidth and performance required in AI, as well as a likely faster refresh rate of technology should all see continued networking investment downstream of GPU.
Cloud – hybrid still in demand
- HPE (not owned) hosted a capital markets/AI day where a number of cloud computing/AI announcements were made.
- The main take for us was the cooperation with Amazon AWS around customers using applications across the HP private cloud solution Greenlake and AWS.
- The company also expanded their partnership with Equinix data centers in order for clients to utilise external clouds with its Greenlake private cloud.
- HPE has been successful with building up strong private cloud business with its Greenlake offering and now working hard to integrate it with public cloud solutions.
- Hybrid cloud still seems to be an infrastructure set up of choice (or at least a piece of the puzzle alongside public cloud) for many (large) enterprises.
Gaming – Mario returns
- Nintendo (owned) made significant announcements during its June Direct presentation, including the upcoming release of a new Mario game called Super Mario Bros. Wonder.
- This 2D entry in the popular franchise, which has sold over 800 million units worldwide, is scheduled to launch on October 20 for the Nintendo Switch.
- There were a number of other announcements around new games and remakes of old popular franchises – which is one of Nintendo’s competitive advantages when it comes to long lasting and sustainable IP.
Portfolio view: The gaming sector is still experiencing headwinds in the market, and we are continuing to see that companies with only really strong titles/releases are the ones that continue to deliver in the current market.
For weekly insights on the latest market updates, please subscribe to our Tech Thoughts podcast.
For more information about the latest trends and forecasts, please visit our official Tech Thoughts page.
We provide investors with access to category leading technology companies, globally. Our assets under management have a total value of more than $1bn, and our limited partners include institutions, family offices and entrepreneurs. Learn more about our funds here.
For enquiries, please contact:
Inge Heydorn, Partner, at email@example.com
Jenny Hardy, Portfolio Manager, at firstname.lastname@example.org
Nejla-Selma Salkovic, Analyst, at email@example.com
About GP Bullhound
GP Bullhound is a leading technology advisory and investment firm, providing transaction advice and capital to the world’s best entrepreneurs and founders. Founded in 1999 in London and Menlo Park, the firm today has 13 offices spanning Europe, the US and Asia. For more information, please visit www.gpbullhound.com.