الأربعاء، 9 فبراير 2022

NEWS TECHNOLOGIE

Late on Monday, Nvidia’s $66B attempt to purchase ARM from Softbank collapsed. Regulators in the US, UK, and EU all raised concerns over the impact of allowing one of ARM’s customers to purchase the company. Had the deal gone through, Nvidia would have been at the center of a silicon empire that dwarfs anything Intel and AMD have ever built.

Nvidia has already telegraphed that it would be fine whether it bought ARM or not. Last year, Nvidia announced its upcoming “Grace” data center CPU based on a future ARM Neoverse core. Announcing the chip before the ARM acquisition went through was a way for Nvidia to telegraph confidence in its own roadmap, no matter what the outcome.

Nvidia’s core graphics business never depended on the ARM acquisition and won’t be greatly harmed by its failure. But what about ARM itself?

ARM Preps for an IPO

We know a few things already. ARM’s CEO, Simon Segars, is stepping down to be replaced by Rene Haas, former head of the company’s intellectual property unit. ARM has also publicly announced that it will begin preparing for an IPO (Initial Public Offering).

If ARM goes through with these plans it’ll probably be the largest chip IPO in history. When SoftBank bought ARM it paid $31.4 billion. Nvidia’s deal was worth $66 billion when it fell through. ARM dominates both the IoT and mobile markets with very little competition. RISC-V is a potential competitor in IoT and low-power devices but scarcely a daunting threat.

According to the regulatory documents ARM filed back in December, the company sees the situation a little differently. In its filing, ARM detailed its view of the competitive situation. The document is an advocacy piece, but a useful one.

ARM’s recent financial performance according to SoftBank’s annual report.

The crux of ARM’s argument is that the mobile market is saturated already. Breaking into the data center and consumer PC markets against a pair of entrenched competitors is anything but easy. Server vendors are conservative. It can take years to bring software projects to fruition. IoT shipments represent a substantial long-term profit center, but an IoT CPU sale earns ARM a fraction of the revenue than a desktop or laptop sale earns Intel or AMD. Lower margins have been key to ARM’s business strategy, but they also limit its revenue growth.

Future Risks

ARM points out that the success of companies like Apple and Amazon does not necessarily benefit ARM itself. Many of ARM’s most successful licensees have architectural licenses, not hard IP licenses. The CPU cores inside an iPhone support the ARMv8 ISA, but they are built with Apple IP. Apple doesn’t contribute design elements from its A-series or M-series chips back to ARM for inclusion in the next round of Cortex CPUs. ARM is also concerned about its ability to retain top-drawer engineering talent in the face of stiff competition from some of its own customers, in addition to the likes of Intel and AMD.

ARM doesn’t argue that it faces critical risks to its core competencies. Its point was more subtle. Faced with entrenched x86 competition, an independent ARM was more likely to focus on mobile and IoT products.

All of that having been said, ARM is still worth somewhere between $30 – $60 billion. Its GPU and CPU cores power billions of devices worldwide. The big question for ARM is whether it should continue to chase the data center and x86 PC markets or not? This is not necessarily easy to answer. ARM licenses its own Cortex CPU designs to many customers. It builds SoC designs for target markets based on customer feedback.

Watch the Datacenter Space

Since 2018, ARM has marketed its server and infrastructure chips under the Neoverse brand. CPUs like the Neoverse N1 are used in Amazon’s Graviton2. Last year, ARM launched the Neoverse V1 for HPC workloads and the Neoverse N2, which ships in Marvell’s Octeon 10 400Gbe PCIe Gen 5 DPUs. ARM isn’t going to just walk away from high-end Neoverse parts, but watch the company’s long-term announcements.

If ARM intends to keep pushing into data centers, it’ll keep updating its Neoverse roadmap with higher-end HPC and server chips. The Neoverse brand covers everything from HPC to edge deployments and 5G infrastructure, so ARM doesn’t have to kill the whole brand to quietly halt research in a given segment. If the company decides to draw down its investment in this area — and its own filing suggests it might — it may happen gradually and probably not right away.

We want to stress that we don’t necessarily expect this outcome. ARM may well decide to ramp up investment in the data center side of its business. But based on the company’s previous statements, there’s a chance it will retrench and focus investment in its most profitable segments rather than ramping its own silicon designs into markets where it currently faces tougher odds.

Now Read:



from ExtremeTechExtremeTech https://ift.tt/VKEm9wR

ليست هناك تعليقات:

إرسال تعليق