Semiconductor deals have reached a frenzy. With more than $100 billion worth of acquisitions announced this year, including two major ones this past week, the chip sector is being rethought and remade by the companies rich enough to strike deals.
The buying frenzy has as much to do with physics as balance sheets. As the billions of transistors that power today’s chips grow closer to a single atom in size, the research and development costs have mounted, steadily increasing to the point where smaller businesses can’t afford the leap to next-generation chips. More than ever, size matters.
“By not addressing your scale, you’re taking a lot of risks, getting yourself into trouble further down the line,” says Pierre Ferragu, an analyst with New Street Research.
The need for scale is one of the arguments that
Advanced Micro Devices
(ticker: AMD) CEO Lisa Su made to Barron’s this past week, after her company reached a deal to acquire
(XLNX) for $35 billion. The all- stock transaction, announced Tuesday, values Xilinx shares at $143, roughly a 25% premium to the previous day’s ending price. AMD expects the acquisition to close by the end of next year.
“The semiconductor industry is a high-stakes investment game,” Su said. “From that standpoint, scale is very important. The more scale you have, the broader the solutions you can put together, the more you can do for your customers.”
But with great scale, comes great responsibility—and risk. For years now, AMD has been a fairly straightforward business—building increasingly powerful computer and graphics processing units for computers, videogame consoles, and data centers. Not a simple enterprise, by any means, but one with a clear objective that provided a reason to buy the stock. AMD has executed fabulously well.
Analysts expect the company to generate $9.4 billion in sales this year, up 40% from the total a year ago, with net income more than doubling to $1.5 billion, or $1.22 per share, on an adjusted basis. Analysts forecast earnings to grow another 50% next year. The growth helps explain why AMD shares have soared 126% in the last 12 months.
The Xilinx deal, though, brings AMD into the unknown. Xilinx has long been known for its expertise in so-called field-programmable gate arrays (FPGAs), chips that can be reprogrammed after they are produced. FPGAs are especially useful in emerging technology, such as 5G wireless infrastructure, for which there aren’t yet custom-designed semiconductors. AMD’s chief rival
(INTC) made a similar bet on FPGAs when it acquired Altera for $16.7 billion in 2015. That hasn’t exactly been a game changer. Shares of the chip giant have returned just 48% over the past five years, badly trailing the
“My first reaction is that AMD is simply following the playbook that Intel has already laid out,” says Brian Bandsma, a portfolio manager at Vontobel Asset Management, says of the Xilinx purchase.
Xilinx sales are projected to shrink 3% this fiscal year, to $3.08 billion from $3.16 billion a year ago. “It doesn’t seem to be a huge growth story right now,” Bandsma says.
One institutional investor tells Barron’s that its fund unloaded its AMD position Tuesday because the stock’s rich multiple—the shares fetch 46 times estimated earnings for the next 12 months—no longer makes sense with Xilinx attached. The growth story has been muddied by non-core businesses, this investor says.
Sure enough, AMD shares fell 4% on the deal news.
“I think we’ve given them a lot of information this morning, so there’s a little bit of digestion,” Su told Barron’s right after the announcement.
If anyone can make the AMD-Xilinx pact work, it’s Su. An engineer by training, she became AMD’s CEO in 2014. Since then, AMD stock has soared 2,200%, compared with a gain of 270% for the
PHLX Semiconductor index.
Barron’s named Su to its list of the world’s best CEOs in 2019.
“We’ve spent the last five years building an execution machine, and that’s not going to change,” Su says when asked about plans to bring Xilinx into the fold.
“We actually don’t need to do this,” she adds. “The base business is going very well, but in this industry you have to think longer-term, and having access to all of this technology capability and market exposure strengthens AMD for the longer term.”
Marvell Technology Group
(MRVL) announced that it is buying fiberoptic-component and data-center chip maker
(IPHI) for $10 billion. The cash and stock transaction values Inphi stock at $157.83 a share—a rich 42% premium to Wednesday’s closing price.
Investors weren’t entirely sold on this deal either. Marvell shares fell 3.3% on the day of the announcement. CEO Matt Murphy told Barron’s that the price is justified, given that Inphi is growing sales at 40% a year.
Here, too, scale is a big part of the rationale, but in a different way. While AMD is looking for new markets, Marvell is doubling down on its existing business, which focuses on data-center chips and wireless infrastructure products. Inphi sells complementary products to many of Marvell’s customers, and its sales are expected to grow 87%, to $682.6 million, this year.
“From a financial point of view, this thing has a kick-butt-looking P&L,” Murphy says of Inphi’s profit and loss statement. “It’s got a lot of revenue growth.”
He adds that scale matters for Marvell because it is using more people and more expensive tools and spending more on manufacturing as it moves to smaller transistors.
“The average cost of a semiconductor advanced node project today is at least $100 million, but if you went back 10 or 15 years ago, it was probably around $10 million,” Murphy says.
“The cost of development and the complexity has gone up so much, that you just have to have enough scale,” he says. “If you invest $100 million in making a chip, you had better sell $500 million of that chip to the right people.”
Write to Max A. Cherney at firstname.lastname@example.org