Ampere Receives A $6.5 Billion Exit
Earlier this week, Ampere announced that it had been acquired by Softbank, the Tokyo-based conglomerate with deep investments across semiconductors, AI and robotics, telco and more. This ends the startup’s independent quest to drive Arm-based silicon innovation deep into the data center and edge and opens a new chapter for the 1,400-member team as part of a much larger and more powerful entity. With the acquisition, Oracle and Carlyle, Ampere’s lead investors, have sold their stakes in the company.
At its launch, Ampere delivered a shot across the bow of x86 dominance in the data center landscape introducing large core count processing for data center workloads based on the Arm core. Their focus was on addressing the balance of performance and efficiency at scale, modernizing an architecture that was designed specifically for the cloud. This has played out in the market with mixed results, with highs of Ampere gaining traction within OCI (note Oracle investment) as a wedge against x86 alternatives. More recently, Ampere had seen sales slump from over $46 million in 2023 to just over $14 million in revenue in 2024 as data center operators attention shifted to accelerated computing to fuel AI, and a value proposition of large quantities of efficient cores became less urgent.
$6.5 Billion – That's a Lot of Cores
When looking at this deal, we need to ask why Softbank, a very savvy investor group, spent $6.5 billion for a company generating $14.6 million in revenue. The answer likely lies in a few strategic broader trends in the market, that trace from the world's governments to this week’s GTC conference, and to the limitations of savvy microprocessor designers vs. industry demand.
At GTC this week, Jensen Huang made clear that the unrelenting demand for computing is growing at an exponential rate. Hitting a nail in the threat of new levels of efficiency of LLMs like Deep Seek, Jenson clarified that these reasoning models require a stunning 100X compute power vs previous forecasts to deliver to their market opportunity. This compute demand will, of course, in part be satisfied by NVIDIA GPUs like the upcoming Blackwell Ultra and Vera Rubin models. But customers are getting increasingly nervous about continued “one basket” investment with team green. We’ve commented on this recently in our coverage of late on Cerebras, the high-powered AI ASIC startup that is delivering some breakthrough performance, and we expect to see a groundswell of growth across accelerator alternatives as large players start chipping away at that 100X capacity demand.
What customers may have this on their radars? Well, a first stop may be looking at the recent announcements of governments stepping aggressively into the AI game, most notably with the Stargate (U.S. government initiative announced earlier this year). We’d also look at something Softbank announced themselves, Cristal Intelligence, a bold move for AI adoption in the Japanese market. Both deals featured OpenAI collaborations. Both also had another highly featured vendor...Arm. Arm and Softbank have a long history together with Softbank maintaining 90% control of the company’s equity. Arm was also notable in the Stargate announcement, and many took their presence as a reflection of their core integration into Grace Blackwell platforms, bringing an efficient head node core in balance with Blackwell GPUs. The company was more prominently featured within the Cristal Intelligence announcement being called out as a strategic innovator for the new “SB Open AI Japan” joint venture that is seeking to unleash agentic computing to automate over 100 million workflows for Softbank group companies.
And agentic computing takes us to our third topic – a dramatic shortage of talented semiconductor designers within the industry to serve this 100X ramp. In fact, this was a key theme at Synopsys’ Executive Forum this week, where the firm rolled out their strategy for agentic semiconductor engineering assistance to help the industry fill this gap. We’ll be delivering more on this story in the coming days, but needless to say, with demand to shrink component designs from historic windows of 7 years down to one year from concept to deployment, the semiconductor arena requires an infusion of engineering cycles as the entire industry innovates the semiconductor design workflow.
With Ampere, Softbank gains just that with a 1,100 member CPU design team to match their architecture design team within Arm, filling what likely is a strategic talent gap for path to market. This marries perfectly with the large customer collaborations rolling out at rapid pace and potentially blends with the much rumored direct from Arm product portfolio expected by the industry later this year.
So what’s the TechArena take?
We’ve been covering Ampere since the foundation of our platform and are excited to see the next chapter for the company and its leadership. I, for one, am intrigued by how the two firms will work under one Softbank umbrella. And, most of all, I’m led into conjecture space where I see the mounting pressure for more competition in the accelerator arena. Reading through the ambitions of the Cristal Intelligence initiative, I see Arm being positioned for much more than efficient core delivery for this landscape. Arm’s CEO, Rene Haas, has a history in GPU product oversight in the seven-year tenure at NVIDIA preceding his move to take the helm at Arm. He’s tapped key talent from NVIDIA to help him lead that charge. Now we’ve got Renee James, a very savvy semiconductor executive with decades of experience in understanding both microprocessor delivery and full stack optimization at Intel dating back to her time as Andy Grove’s protege, growing what seems to be tighter alignment with team Haas.
Does this deliver an opportunity for tech innovation that we have yet to imagine? We can’t wait to find out what comes next.
Earlier this week, Ampere announced that it had been acquired by Softbank, the Tokyo-based conglomerate with deep investments across semiconductors, AI and robotics, telco and more. This ends the startup’s independent quest to drive Arm-based silicon innovation deep into the data center and edge and opens a new chapter for the 1,400-member team as part of a much larger and more powerful entity. With the acquisition, Oracle and Carlyle, Ampere’s lead investors, have sold their stakes in the company.
At its launch, Ampere delivered a shot across the bow of x86 dominance in the data center landscape introducing large core count processing for data center workloads based on the Arm core. Their focus was on addressing the balance of performance and efficiency at scale, modernizing an architecture that was designed specifically for the cloud. This has played out in the market with mixed results, with highs of Ampere gaining traction within OCI (note Oracle investment) as a wedge against x86 alternatives. More recently, Ampere had seen sales slump from over $46 million in 2023 to just over $14 million in revenue in 2024 as data center operators attention shifted to accelerated computing to fuel AI, and a value proposition of large quantities of efficient cores became less urgent.
$6.5 Billion – That's a Lot of Cores
When looking at this deal, we need to ask why Softbank, a very savvy investor group, spent $6.5 billion for a company generating $14.6 million in revenue. The answer likely lies in a few strategic broader trends in the market, that trace from the world's governments to this week’s GTC conference, and to the limitations of savvy microprocessor designers vs. industry demand.
At GTC this week, Jensen Huang made clear that the unrelenting demand for computing is growing at an exponential rate. Hitting a nail in the threat of new levels of efficiency of LLMs like Deep Seek, Jenson clarified that these reasoning models require a stunning 100X compute power vs previous forecasts to deliver to their market opportunity. This compute demand will, of course, in part be satisfied by NVIDIA GPUs like the upcoming Blackwell Ultra and Vera Rubin models. But customers are getting increasingly nervous about continued “one basket” investment with team green. We’ve commented on this recently in our coverage of late on Cerebras, the high-powered AI ASIC startup that is delivering some breakthrough performance, and we expect to see a groundswell of growth across accelerator alternatives as large players start chipping away at that 100X capacity demand.
What customers may have this on their radars? Well, a first stop may be looking at the recent announcements of governments stepping aggressively into the AI game, most notably with the Stargate (U.S. government initiative announced earlier this year). We’d also look at something Softbank announced themselves, Cristal Intelligence, a bold move for AI adoption in the Japanese market. Both deals featured OpenAI collaborations. Both also had another highly featured vendor...Arm. Arm and Softbank have a long history together with Softbank maintaining 90% control of the company’s equity. Arm was also notable in the Stargate announcement, and many took their presence as a reflection of their core integration into Grace Blackwell platforms, bringing an efficient head node core in balance with Blackwell GPUs. The company was more prominently featured within the Cristal Intelligence announcement being called out as a strategic innovator for the new “SB Open AI Japan” joint venture that is seeking to unleash agentic computing to automate over 100 million workflows for Softbank group companies.
And agentic computing takes us to our third topic – a dramatic shortage of talented semiconductor designers within the industry to serve this 100X ramp. In fact, this was a key theme at Synopsys’ Executive Forum this week, where the firm rolled out their strategy for agentic semiconductor engineering assistance to help the industry fill this gap. We’ll be delivering more on this story in the coming days, but needless to say, with demand to shrink component designs from historic windows of 7 years down to one year from concept to deployment, the semiconductor arena requires an infusion of engineering cycles as the entire industry innovates the semiconductor design workflow.
With Ampere, Softbank gains just that with a 1,100 member CPU design team to match their architecture design team within Arm, filling what likely is a strategic talent gap for path to market. This marries perfectly with the large customer collaborations rolling out at rapid pace and potentially blends with the much rumored direct from Arm product portfolio expected by the industry later this year.
So what’s the TechArena take?
We’ve been covering Ampere since the foundation of our platform and are excited to see the next chapter for the company and its leadership. I, for one, am intrigued by how the two firms will work under one Softbank umbrella. And, most of all, I’m led into conjecture space where I see the mounting pressure for more competition in the accelerator arena. Reading through the ambitions of the Cristal Intelligence initiative, I see Arm being positioned for much more than efficient core delivery for this landscape. Arm’s CEO, Rene Haas, has a history in GPU product oversight in the seven-year tenure at NVIDIA preceding his move to take the helm at Arm. He’s tapped key talent from NVIDIA to help him lead that charge. Now we’ve got Renee James, a very savvy semiconductor executive with decades of experience in understanding both microprocessor delivery and full stack optimization at Intel dating back to her time as Andy Grove’s protege, growing what seems to be tighter alignment with team Haas.
Does this deliver an opportunity for tech innovation that we have yet to imagine? We can’t wait to find out what comes next.
