There may be changes in the artificial intelligence chip space.
Until a few days ago, it appeared that AI hyperscalers had only a few options to choose from when equipping their data centers with parallel processors. Nvidia’s graphics processing units (GPUs) have been the most popular choice by far, and amdGPUs have also captured a share of the market. Another option that the largest data center operators adopted was to partner with broadcom (AVGO 4.15%, To develop custom AI accelerators – called application-specific integrated circuits – that are optimized for a particular type of workload.
However, recently news came that alphabet (GOOG 1.07%, (Google 1.18%, May signs a deal to sell a large quantity of its tensor processing units (TPUs), which it collaborated with Broadcom to design. meta platformSuch a deal could herald the arrival of a new competitor in the computing hardware market, which is why buying these three stocks now might make more sense than adding shares of Nvidia,
Image Source: Getty Images.
Three companies are sure to benefit
The news article regarding the possibility that Meta would purchase TPU from Alphabet acknowledges that this is just a possibility. The companies have made no concrete public comments on the matter, nor has any clear dollar figure been released, although it has been reported to be in the billions of dollars. With all that said, there are three companies in the TPU value chain, and I think each of them is worth a buy right now.
The most obvious company that would benefit is Alphabet.
Alphabet gets most of its revenue ($74 billion of $102 billion in the third quarter) from advertising. Advertising is not a stable business – it rises and falls depending on the strength of the consumer and the willingness of businesses to spend on advertising. The company also has a cloud computing business that is thriving due to the general shift toward cloud workloads and new artificial intelligence workloads coming online.
,-1.18,-3.77
current price
,316.41
key data points
$3864B
day limit
,314.49 ,319.74
52wk range
,140.53 ,328.83
volume
742K
average volume
38m
gross margin
59.18,
dividend yield
0.26,
Until now, if companies wanted to use TPUs for their processing needs, their only option was to rent time on them through Google Cloud. They couldn’t buy anything. If Alphabet starts selling its TPUs, it would open up a new source of revenue that investors haven’t yet accounted for. That’s why Alphabet’s stock rose by double-digit percentages a few days after news broke about a potential Meta deal.
Broadcom would also be a big beneficiary if these sales go through, as Alphabet pays the chip designer for each TPU it purchases. It’s unknown whether Broadcom will make more money if Alphabet decides to sell them to third parties, but custom AI chips have become a big part of Broadcom’s business.
In the third quarter of fiscal 2025, which ended Aug. 3, $5.2 billion of Broadcom’s $15.9 billion in revenue came from AI-related sources. However, this divide is growing rapidly; Management expects $6.2 billion in AI revenue in the fiscal fourth quarter.
,-4.15,-16.72
current price
,386.24
key data points
market cap
$1903B
day limit
,385.10 ,395.36
52wk range
,138.10 ,403.00
volume
728K
average volume
24 months
gross margin
63.13,
dividend yield
0.59,
If the Alphabet and Meta deal goes through, that number could skyrocket in 2026, boosting Broadcom’s financial position accordingly.
The last company to benefit from this deal is one of my favorites in the AI field: Taiwan Semiconductor (TSM 0.99%,,
today’s change
,-0.99,-2.88
current price
,288.63
key data points
market cap
$1512b
day limit
,286.48 ,290.89
52wk range
,134.25 ,311.37
volume
313K
average volume
13m
gross margin
57.75,
dividend yield
0.99,
Taiwan Semiconductor benefits no matter who is the ultimate winner
Nvidia has been the clear leader in the AI accelerator field up to this point, but its dominance may be slipping. However, all of these AI hardware providers are “fableless” chip manufacturers. They design the chips, but they don’t manufacture them in-house. Instead, they outsource that work to chip manufacturing companies, and the largest of these companies by a wide margin is Taiwan Semiconductor.
This establishes Taiwan Semiconductor as a neutral (but important) player in the AI world. As long as the expenditure on AI chips increases, its business will grow. If Nvidia’s dominance wanes and Broadcom’s custom AI units take off, TSMC stands to benefit equally.
Regardless of the company’s ability to deliver profits as demand for its AI chips grows, TSMC’s stock trades at a much smaller premium than the other stocks discussed here.
GOOGL PE Ratio (Forward) data by YCharts.
Trading at 27.5 times forward earnings, TSMC shares are a great deal, and retail investors should buy them along with Alphabet and Broadcom, especially if Alphabet starts selling its TPUs to other data center operators.
I still think Nvidia remains a strong investment choice, but if TPU becomes available for widespread purchase, all three could become better buys.
<a href=