Of all the cloud giants, Microsoft (MSFT +1.69%) has been thought to be the leader in artificial intelligence.
But one sell-side analyst thinks others have caught up. Furthermore, the analyst thinks of all three major cloud providers, which is where the bulk of AI training takes place, Microsoft is the most beholden to Nvidia (NVDA +1.80%) and its high-priced chips.
So, should investors be concerned?
NASDAQ: MSFT
Key Data Points
D.A. Davidson downgrades Microsoft to "neutral"
In late September, D.A. Davidson analyst Gil Luria downgraded Microsoft from buy to neutral, while leaving his price target unchanged at $475, or around 8% higher than the $410 stock price today.
Downgrading Microsoft is rare, with the vast majority of Wall Street analysts having a Buy rating on shares. After all, Microsoft has lots going for it. It has not one but several high-growth and high-margin software businesses, a booming cloud computing platform, and other high-margin franchises in social media, gaming, and PC operating systems. Moreover, Microsoft has been at the forefront of the AI revolution, with its early investment in OpenAI, the parent company of ChatGPT, giving Microsoft exclusive access to OpenAI's leading generative AI models.
The early lead, combined with Microsoft's financial strength -- it has a AAA bond rating, higher than the U.S. government, after all -- has afforded the company a relatively high valuation at just under 35 times earnings.
That valuation is justified if Microsoft can continue leading in AI. However, Luria sees Microsoft's early leadership eroding, with the company also having a higher cost structure than other clouds.
Is Microsoft's AI lead shrinking or gone?
Luria's downgrade is predicated on two things. First, he thinks both Amazon (AMZN 1.40%) and Alphabet (GOOG +0.05%) (GOOGL 0.04%) have caught up in terms of AI-infused solutions.
That of course is open for debate. OpenAI still appears to be the go-to leader in the generative AI LLM space. That being said, Amazon's investments in Anthropic and Google's massive investments in its Gemini model may have equalized the competition, at least according to the analyst.
Anthropic is the company that produces the Claude LLMs, and is run by Dario Amodei, who was previously the VP of Research at OpenAI. This summer, the company posted some impressive performance benchmarks against ChatGPT4.0, which was the leading model at the time. In addition, Google has also made claims its Gemini model outperforms ChatGPT on a number of benchmarks, while also being updated more frequently.
While Microsoft was out to an early lead, Luria seems to think the gap between Anthropic, Gemini, and ChatGPT has closed. He points to Amazon now adding as much new business to its cloud offering as Microsoft last quarter, and Google Cloud's accelerating growth rate matching Microsoft Azure's last quarter as well.
Image source: Getty Images.
And with more advanced homegrown AI accelerators
Besides catching up to ChatGPT, Luria also thinks Amazon and Google have a cost advantage, because of their early investments in their own AI accelerators. Google started earliest of all, developing its Google Tensor Processing Units (TPUs) in 2015 and deploying them in 2018. Amazon also began designing its own processors in 2015 after acquiring Annapurna Labs, and developed its Trainium AI chips in 2020.
The ability for Google and Amazon to execute workloads on its own in-house accelerators is a massive cost advantage, since Nvidia chips are so expensive. However, since Microsoft has been late to the chip development game, Luria believes Microsoft now has a cost disadvantage, given that Nvidia GPUs are so expensive.
According to a Bloomberg analysis, Microsoft accounts for nearly 20% of Nvidia's revenue, whereas Alphabet and Amazon account for only about 5% each. That's astonishing especially for Amazon, whose cloud computing unit AWS is actually larger than Microsoft's.
Given the concerns around the return on capital being spent on Nvidia chips, this is why Luria thinks Microsoft may not get as high a return as the competition.
The counter-argument
I don't think that investors were under any illusions that there would be a lot of competition in the AI space. After all, Microsoft has well-funded competitors, and it has for many years. Moreover, ChatGPT just released OpenAI o1, its latest model that OpenAI says can "reason" by taking a longer time to think about questions and prompts, leading to an ability to answer more complex problems. So, it's possible that as Anthropic and Gemini pull even, ChatGPT surges one step ahead, maintaining a slight lead.
Moreover, Microsoft has begun designing its own chips, called Maia, last year. While Microsoft is a good five to eight years behind Amazon and Google in designing chips, it's unclear whether that is a sustainable lead. After all, these three will generally have their chips produced on similar foundry processes, so it's hard to say which chip will ultimately be more efficient. So while Microsoft is "behind" as of now, as Anthropic and Gemini were to OpenAI, there doesn't seem to be a reason why Microsoft couldn't make effective internal accelerators going forward and catch up.
The gist is that the AI races are very early, and all three major clouds should have competitive AI models with lower-cost internal accelerators for workloads that can be transferred away from Nvidia chips.
As all three benefited from cloud computing because of their hard-to-match capital and technical ability, Microsoft, Amazon, and Google seem set to each benefit from the AI revolution as well. All three remain solid buys today.
