The intersection of synthetic intelligence (AI) and semiconductor expertise is producing some of the most dynamic shifts in the tech trade at present.

On that observe, three Motley Fool contributors who cowl tech and AI shares put their heads collectively to spotlight their prime inventory picks amongst AI-focused semiconductor shares at present. They got here up with reminiscence expertise frontrunner Micron Technology (MU -3.91%), trade legend Intel (INTC -5.16%), and AI chip design chief Broadcom (AVGO -2.78%).

Read on to evaluate how these chip titans profit from the ongoing AI surge.

Intel goals to catch up to Nvidia and TSMC unexpectedly

Billy Duberstein (Intel): Intel has badly lagged different AI firms in latest years and in 2024, with the inventory down over 25% on the yr to this point as the Nasdaq has rallied virtually 8%.

That’s considerably warranted — in any case, Intel is presently in a multiyear try to catch up to Taiwan Semiconductor Manufacturing in phrases of its means to produce modern chips. Success wouldn’t solely imply Intel’s designed chips would obtain market management as soon as once more, however that Intel would additionally turn out to be a profitable and worthwhile foundry for third events.

Achieving course of expertise management and successful new shoppers is an extended, powerful slog, however Intel has proven good progress over the previous yr or so, hitting milestone after milestone on its forecast timeline to obtain 5 chip nodes in 4 years.

And simply final week, Intel unveiled its new Gaudi 3 chips, the accelerators Intel hopes will compete with Nvidia GPUs. In a shock, Intel’s administration claimed the new Gaudi chips outdo the present Nvidia H100 on a number of key metrics. While Nvidia is already transferring on to its new Blackwell chips later this yr, Gaudi has the potential to discover some small share of the AI market, which is popping out to be a really massive market.

Aside from sheer {hardware} chops, Intel can also be main massive consortiums aiming to kind open supply options to each AI programming software program, taking intention at Nvidia’s CUDA software program moat, in addition to AI ethernet networking, difficult Nvidia’s Infiniband answer.

Intel additionally simply broke out its merchandise and foundry phase financials individually, giving traders a way of what Intel could also be producing in earnings with out the burden of all its foundry capital. Outside of Intel’s foundry, its chip merchandise made about $12.4 billion in working revenue final yr. That’s not too unhealthy, contemplating Intel has fallen behind on course of expertise in latest years, and that this does not include a lot of any Gaudi income. It additionally means Intel’s valuation will not be demanding, with a market cap of simply $160 billion.

Of course, Intel’s foundry misplaced a whopping $7 billion final yr alone. But that is not fully stunning, given the big capital outlays and depreciation that should be taken forward of Intel bringing in income.

Will Intel pull it off? The firm is investing closely behind the effort, procuring the first high-NA EUV machine from ASML Holdings. Not solely that, however Intel has additionally locked down the majority of ASML’s high-NA EUV machines for this yr — an aggressive funding to try to pull forward of TSMC in 2025.

For these in search of AI winners that have not but taken off, Intel may very well be a contrarian play. However, it is also riskier, and far more depending on stable execution. Still, success over the subsequent couple of years could lead on to significant upside for the inventory — most likely greater than the others which have already run.

Micron is a good AI wager, regardless of who wins the accelerator wars

Anders Bylund (Micron Technology): Every chip designer value its sodium silicate is arising with highly effective AI accelerator chips as of late. If you’ll be able to’t beat the uncooked efficiency of Nvidia’s H100 and Blackwell chips, possibly you’ll be able to provide a extra power-efficient answer, or a less expensive chip, and so forth. There’s sufficient demand for AI methods and options to create all kinds of exploitable area of interest markets.

Micron Technology takes a special method to the AI alternative.

As a number one supplier of reminiscence {hardware}, Micron advantages from the AI-driven growth in high-performance computing. Every high-end number-crunching machine wants a ton of quick DRAM reminiscence, and most additionally depend on massive arrays of solid-state storage units (SSDs) constructed round NAND reminiscence chips.

The reminiscence sector as an entire is on the rise, however Micron goes a pair of steps additional.

• Right now, chief rivals Samsung and SK Hynix haven’t got a solution to the energy effectivity of firm’s newest high-bandwidth reminiscence (HBM) merchandise. This ought to make Micron the best option for conditions the place a really great amount of high-speed DRAM reminiscence is required — comparable to the supercomputers that prepare AI engines with the assist of accelerators from Nvidia and its rivals.
• In reality, Nvidia’s Blackwell AI accelerators are transport with up to 192 gigabytes of Micron HBM per module. In different phrases, Micron rides Nvidia’s coattails in an extremely literal means.

Now, Micron traders have seen these progress catalysts coming in and the inventory trades only a hair beneath all-time highs. Micron’s valuation ratios additionally run excessive. If you are in search of a low-priced low cost deal, this is not the inventory for you.

But it is a completely different story should you consider that the reminiscence market is altering on a deeply basic degree, putting Micron in a golden age of expertise management amid a game-changing AI growth. In this mild, Micron’s excessive valuation would possibly simply be “the new normal” with the potential of additional acquire over the subsequent couple of years.

So it’s best to evaluate your tolerance for market danger and act accordingly. For growth-oriented traders with a strong danger urge for food, Micron presents a direct funding in the AI frenzy with out having to choose a winner in the accelerator chip wars.

Big tech AI chips are actually all about this prime semiconductor inventory

Nicholas Rossolillo (Broadcom): It’s been a busy week in AI and semiconductors, with Alphabet asserting a brand new Arm Holdings-based processor. Google additionally mentioned how its newest Tensor Processing Units (TPUs) are coaching Google AI. Meta Platforms introduced its newest customized AI chip for processing some of its information middle computing. And even Intel tried to get in on the motion with the debut of its newest Gaudi 3 AI chip.

At this level, all of the tech giants that make up the “Magnificent Seven” have, at the very least, hinted that they are engaged on one thing in the identify of competing with Nvidia.

In all actuality, these firms aren’t a lot making an attempt to compete with Nvidia as a lot as they’re in search of a means to avoid wasting cash.

At any price, what few traders understand is that the majority of these massive tech chip designs aren’t precisely of their very own making. They even have to faucet the semiconductor trade itself in order to design these AI methods, which they’re ready to accomplish that since they’ve such ample monetary assets at their disposal.

The largest customized AI processor design home at this level is none aside from chip titan Broadcom, which is answerable for serving to Google with its customized TPUs, in addition to doing the heavy lifting design work for Meta, and an as-yet unnamed third tech large on one other AI mission.

As a end result, some 35% of Broadcom’s 2024 semiconductor gross sales are anticipated to be from AI.

These days, the market is targeted on how effectively Broadcom can combine and unlock profitability from its mega-acquisition of cloud software program important VMware. But the AI chips may very well be an actual long-term recreation changer. At 28 occasions the present fiscal yr anticipated earnings per share, Broadcom inventory may really nonetheless be an honest worth for the lengthy haul.