NVIDIA (NVDA -1.19%) this week agreed to buy chip designer Arm Holdings from SoftBank Group (SFTB.Y -0.60%) in a deal worth up to $40 billion. The deal won't include Arm's Internet of Things (IoT) service group, which will be spun off into new units within SoftBank.
NVIDIA will initially pay SoftBank $21.5 billion in stock and $12 billion in cash. NVIDIA will issue an additional $1.5 billion in shares to Arm's employees, then pay up to $5 billion in cash or common stock to SoftBank if Arm achieves certain financial targets.
The deal has been approved by the boards of directors at NVIDIA, SoftBank Group, and Arm, and is expected to close within 18 months. NVIDIA expects the takeover to be "immediately accretive" to its non-GAAP gross margin and EPS.
The deal isn't surprising, since the rumors had been circulating since late June, but it's still a game-changing deal that could significantly expand NVIDIA's business beyond GPUs and threaten its rival chipmakers Intel (INTC 0.61%) and AMD (AMD 0.27%).
What does Arm Holdings do?
Arm's low-power chip designs are found in over 95% of the world's smartphones, but the company doesn't produce any chips on its own. Instead, it licenses its designs to other chipmakers, including NVIDIA, Qualcomm (QCOM -0.97%), and Apple (AAPL -2.12%), and collects royalties and licensing fees.
Arm's revenue rose 2% to 206.7 billion yen ($1.94 billion) in fiscal 2019, which ended on March 31, but it posted a net loss of 42.8 billion yen ($400 million), compared to a profit of 134 billion yen ($1.26 billion) a year earlier. Excluding a one-time gain from setting a joint venture in China in 2018, Arm would have remained unprofitable over the past three years, due to slower smartphone sales, the trade war, and COVID-19 disruptions.
Why did NVIDIA buy Arm?
Arm's recent growth has been weak, but integrating its operations with NVIDIA's could eliminate redundancies, cut costs, and boost its profits. Integrating Arm's full business would boost NVIDIA's annual revenue by over 10%, and its expectation for the deal to "immediately" boost its earnings suggests that sweeping layoffs are right around the corner.
But afterward, NVIDIA will start earning ARM's royalties and licensing fees from rivals like Qualcomm, which previously drove its ARM-based Tegra CPUs out of the mobile chipset market with its Snapdragon SoCs. Much of Qualcomm's past success in the mobile market can arguably be attributed to its controversial practice of bundling SoCs with essential wireless licenses.
NVIDIA might implement a similar strategy by bundling its Tegra CPUs, which generated 13% of its revenue last year, with Arm's licenses. Since NVIDIA would no longer need to pay licensing fees to Arm, it could also sell its ARM-based chips at lower prices than competitors like Qualcomm.
That aggressive strategy could lock in Tegra's core customers across the automotive, robotics, drone, and gaming console markets, and potentially support its return to the mobile market or Tegra's expansion into data centers (which NVIDIA currently serves with its high-end GPUs).
However, NVIDIA can't simply cut Qualcomm, Apple, or other chipmakers off from Arm's licenses to sell its first-party chips for three reasons: Those companies are all locked into multi-year contracts with Arm; abruptly cutting them all off would inevitably spark lawsuits and antitrust probes; and Arm's revenue would dry up.
Why should Intel and AMD pay attention?
Intel and AMD both primarily sell x86 CPUs, which are generally more powerful but less power-efficient than Arm-based CPUs. That's why Intel and AMD hold a duopoly in the PC CPU market, while Intel's Xeon CPUs dominate the high-end data center market.
However, Arm-based chipmakers, especially Qualcomm, are gradually making inroads in the PC and data center markets with more powerful chipsets. Microsoft has also been embracing that shift with ARM-based versions of Windows.
NVIDIA's takeover of Arm could accelerate those efforts with higher-end chip designs for PCs and servers. It could build new Tegra CPUs based on those designs, then bundle them with its market-leading GPUs -- which are often used to run high-end games on PCs and process AI tasks in data centers.
If that happens, Intel's efforts to launch its own discrete GPU, the Xe, could fall flat. Moreover, Intel's PC and data center CPU businesses, which are already under pressure from its ongoing chip shortage and a delay of its 7nm chips, could need to counter AMD and NVIDIA at the same time.
AMD, which already competes with NVIDIA in the GPU market, could also be threatened by NVIDIA's moves into the ARM-based CPU space. AMD has been gaining ground against Intel in the desktop and notebooks markets, but the arrival of new power-efficient Tegra CPUs for notebooks bundled with mobile GeForce GPUs could throttle AMD's progress.
The bottom line
NVIDIA's takeover of Arm will enable it to earn licensing fees for nearly every smartphone sold worldwide, boost Tegra's margins by eliminating licensing fees, profit from Arm's long-term replacement of x86 chips, and possibly herald its return to the mobile, PC, and data center CPU markets. In other words, it could be well worth the $40 billion price tag.