This is not investment advice. The author has no position in any of the listed stocks. wccftech.com has disclosure and ethics policies.
ASICs, focused on custom AI, are increasingly billed as Nvidia’s Achilles heels, and by leveraging the strength in the GPU arena, Nvidia can defeat all the painstaking demand-related buildings. However, Morgan Stanley’s research now believes that ASIC poses no threat to Nvidia’s broader success story.
Morgan Stanley: Despite ASICS, Nvidia will continue to maintain its dominant market share.
The ASIC category is neither superior nor inferior to commercial GPUs. Another way to achieve the same result.
Over the past six months, the momentum of the AI sector has been clear…
– Jukanlosreve (@jukanlosreve) February 14, 2025
Witness, Morgan Stanley Research is currently writing a thorough note on the evolving GPU-versus-ASIC paradigm, claiming that Nvidia is not facing a serious threat to its dominant position in the market .
Titan’s research division on Wall Street initially acknowledged that “the AI sector’s momentum has clearly shifted to custom silicon,” and so is the market. Quarterly revenues are supported by Broadcom (AVGO) at a market capitalization of just $3.2 billion per quarter.
Morgan Stanley Research notes that development budgets for certain ASICs are generally less than $1 billion, often far lower. In contrast, Nvidia spends an insane amount of money to deliver pure white products.
“…Nvidia will invest around $16 billion in R&D this year alone. With that funding, Nvidia will be able to maintain a four-to-five year development cycle by operating three design teams one after another. Plus, five Year-round innovation.
Of course, custom ASICs such as Google’s tensor processing units (TPUs) are much more customizable. However, Morgan Stanley claims that “the largest AI training and inference clusters are currently not highly customized,” and Nvidia continues to optimize GPUs for transformer models. Going forward, Morgan Stanley believes that the largest utility for highly customized ASICs lies in “Legacy Workloads.”
However, some may argue that the cost factor is the biggest draw for the customized Ashic, with Nvidia’s H100 priced at $20,000, but it costs just $3,000. yeah.
But here, Morgan Stanley argues there are hidden costs to consider. For example, cluster costs for ASICs are often substantially higher because they employ much more expensive optical connection technology, but Nvidia’s 72‑ GPU NVLink domain offers much cheaper copper-based technology. I’m using it.
Furthermore, unparalleled purchasing power allows Nvidia to easily secure a priority rate for high-bandwidth memory (HBM) chips. Titan on Wall Street continues:
“The same applies to Cowos. Many ASICs use smaller dies on a larger stack, so the Cowos costs can be higher than Nvidia’s costs. Of course, Nvidia’s wafer costs are Reticle It could be higher due to the closure die, but overall Nvidia is delivered exceptionally. Value.
Next, Morgan Stanley argues that “software developer time” should also be incorporated into the total cost of ownership (TCO) of an ASIC. Development experience.
Morgan Stanley then went on to argue that “Nvidia and AMD are ASIC competitors (this year), especially in the second half.”
“The size of AMD’s investment across the ecosystem tends to go far beyond that of ASIC vendors. This year, AMD has completed two acquisitions of AI software assets. One of these – the acquisition of ZT systems – Major server ODM acquisitions, ODM business owns such assets in relation to rack and cluster computing, do you replicate this?
Next, Morgan Stanley Research notes that in 2024, commercial silicon controlled 90% of the market leading Nvidia’s $98 billion chip-based revenue (GPU manufacturers still had the last quarter). (On a temporary basis, as it has not announced revenues) AMD has dragged on at $5 billion. In contrast, Custom ASIC managed only 10% of the market in 2024, with Broadcom earnings of $8 billion, with ALCHIP and Marvel taking over, with total revenues of just $2 billion.
The key predictions are: Morgan Stanley declares:
“We expect a slight increase in our 90% share of commercial products this year.”
To bolster this forecast, the Wall Street giant points out that Broadcom is overly dependent on Google’s TPU to promote sales. However, in 2025, Nvidia will “have a 50-100% increase over TPU.”
Similarly, Marvel is increasingly dependent on Amazon’s Trainium Asic. Amazon is set to double ASIC purchases to $4 billion this year, but it is expected that purchases from Nvidia will also be “more than doubled.”
Therefore, Morgan Stanley believes Nvidia’s “revenue momentum for the second half of 2025 will be significantly stronger than builds from ASIC or AMD.”
ASIC growth could explode from 2026 onwards, with Broadcom’s proprietary SAM analysis projecting that “Google, Meta and Bytedance could achieve revenues of $60 billion to $90 billion in fiscal year 2027.” It’s there.
Nevertheless, Morgan Stanley believes that “the total addressable market (TAM) for AI ASICs will grow from $12 billion in 2024 to $30 billion in 2027.
Finally, the investment bank said:
“Nvidia’s biggest short-term risk is US export control, which is equally a problem for AVGOs. In the long run, the biggest risk is not competition but slowing investment.
