By Jeff Brown, Editor, The Bleeding Edge 
On a beautiful fall day in November 2020, Apple announced two new laptop computers and a Mac Mini… They looked more or less the same as their predecessors, so the announcement wasn’t received with much fanfare. But that didn’t mean it wasn’t a big deal. In fact, it was a major shift in computing. Apple broke a 15-year partnership with Intel by replacing its x86 architecture with an ARM-based semiconductor of its own design. Not only did Apple stop using Intel chips, but it started using semiconductors of its own design — manufactured by TSMC (TSM) — and licensing ARM Holdings’ (ARM) semiconductor designs. Most analysts will say that the moment was the beginning of Intel’s downfall. I would argue it started more than two decades earlier. Marketing Success… Product Failure Intel was always more than just a semiconductor company. It has been a prolific marketer for decades. So much so that it became a household name, which is odd for a company with a product that no one ever sees. I’m sure many of us remember the much-hyped line of Pentium processors. I’d bet that most of us have used a PC that had one at one stage or another.  The marketing campaigns may have been wildly successful, but the Pentium was a product disaster. In 1994, a major flaw was discovered that required the chips to be replaced. Intel’s mistake cost the company $475 million to repair. Then, shortly after that mess — in an effort to launch central processing units (CPUs) with faster speeds — the Pentium 4 line of CPUs became famous for excessive power consumption and overheating due to flaws in its semiconductor design. This was the moment that gave Advanced Micro Devices (AMD) the window to gain market share from the semiconductor giant. And AMD has been eating Intel’s lunch ever since. Intel’s Biggest Mistake We all started using laptops more than 20 years ago. It's common sense to all of us using mobile devices that power consumption is important… it impacts our battery life. That’s why power efficiency is so critical. Intel’s x86 architecture was designed in 1978, at a time when every computer was plugged into a fixed source of electricity. There were no mobile devices back then. So it makes perfect sense that the semiconductor architecture should change to adapt to a world of mobile devices. That was Intel’s biggest mistake. And it was ARM’s biggest opportunity, which it capitalized on by designing power-efficient semiconductors optimized for mobile devices. Intel completely missed what is the largest and most explosive trend in electronics history — the rise of smartphones. And most won’t remember this, but Intel had the opportunity to dominate the industry. Apple served that opportunity on a platter for Intel, asking it to design and produce semiconductors for the iPhone. Intel turned Apple down. That’s now a painful memory for those involved at the time. This decision resulted in Intel completely missing out on the multibillion-unit smartphone market opportunity and created the opening for companies like Qualcomm, Apple, MediaTek, and others to develop mobile processors designed on ARM’s power-efficient semiconductor technology. And by the 2010s, Intel’s semiconductor manufacturing problems started to become obvious to the whole industry. For years, it became stuck on the 14-nanometer (nm) manufacturing process while the rest of the industry jumped to 10 nm and then to 7 nm. TSMC, the world’s largest contract manufacturer of semiconductors, was three generations ahead of Intel’s capabilities — years ahead from a competitive standpoint. I was reminded of these blunders in the last couple of weeks, while I was researching Intel’s latest research and development efforts. Intel struggled horribly with the 10 nm and 7 nm manufacturing processes. It was one manufacturing problem after the other, all with years of delays. Then, in an effort to catch up to TSMC and return to a position of manufacturing leadership, Intel placed a massive bet… It Was a Gamble Intel sought to leapfrog ahead of several manufacturing generations… to what it calls 20A and 18A manufacturing processes. The “A” is for angstrom, a unit of measure equivalent to 0.1 nanometers. Therefore, 20A is 2 nanometers, and 18A is 1.8 nanometers. Intel struggled with 20A and scrapped that program due to the costs. It has since gone all in on the 18A manufacturing node. Just last month, Intel proclaimed great progress on the 18A process and gave guidance that it would be manufacturing bleeding-edge semiconductors using this advanced technology in the first half of 2025. And yet, just days ago, Broadcom (AVGO) — the single largest potential customer for Intel’s 18A manufacturing process — announced that it was unsatisfied with Intel’s 18A manufacturing process. Broadcom was very clear that Intel’s 18A process is not yet viable for mass production. It pains me to see one disappointment and failure after another from Intel, one of the most important and successful technology companies in history. It was founded by Gordon Moore — of the famous Moore’s Law. And it proved that his theory of exponential growth in semiconductors was not only proven to be accurate… but had an extraordinary impact on economic growth. It saddens me to say that Intel (INTC) is a company to avoid. What company would want to take the risk of trusting Intel at this stage? Can it deliver on quality and meet its scheduled timelines for a bleeding-edge semiconductor design, after more than a decade of failures? It can be hard to imagine that a company with a brand name like Intel, $52 billion in annual revenues, and $35 billion in cash might be in deep trouble. But when we step back and look at the big picture, it’s much easier to understand. Deep Trouble Intel’s gross margins have collapsed from around 62% in 2017 down to just 40% this year. Annual revenues this year are down 3.5% compared to 2023 despite being in the largest boom in data center infrastructure buildout in history. Worse, Intel sits on $53 billion in debt right now and will have a negative free cash flow of $14.8 billion this year. Further multibillion-dollar losses will continue into next year. From a competitive standpoint, AMD now has 23% of the desktop CPU market, 20.1% of mobile CPUs, and 24.1% of the server CPU markets. And while Intel ships more volume (in terms of units) than AMD… AMD dominates the server market when it comes to high-performance semiconductors. This is easily understood, looking at the numbers: High-performance, bleeding-edge products carry much higher margins. AMD is crushing Intel in this space and its market share will continue to grow. So where does that leave Intel? What does the future hold? Intel needs to decide if it wants to become a legitimate competitor to TSMC as a contract manufacturer of semiconductors for fabless semiconductor companies. Or not. Hard Choices One of the biggest problems is that Intel, like Samsung, has a semiconductor manufacturing arm that is closely tied to its product divisions. Samsung Electronics and other divisions of Samsung that use semiconductors are prioritized, just like Intel’s manufacturing division prioritizes Intel’s semiconductor products. This is a hard sell if you’re looking for someone to manufacture your semiconductor, knowing that you’ll be taking the back seat to their internal customers. AMD actually spun out its own manufacturing arm in 2009 to avoid bankruptcy. It was one of the smartest moves that the company made in its history. AMD’s manufacturing division became known as Global Foundries (GFS) — a much smaller contract manufacturer than TSMC — with only $6.7 billion in annual revenue. After a decade of pain, Global Foundries cleaned up its operations and is financially healthy, generating $1.4 billion in free cash flow this year. But it has nowhere near the capital or scale needed to build bleeding-edge semiconductor manufacturing capabilities, which is why it tends to produce products with more mature manufacturing processes. Intel CEO Pat Gelsinger is going to have to make some very hard decisions about how to stop the bleeding. It’s already rumored that Intel may have to sell off Altera, another semiconductor company that it acquired in 2015 for $16.7 billion. Altera specializes in field programmable arrays (FPGA). Marvell (MRVL), Microchip (MCHP), or Broadcom (AVGO) could be buyers. It’s a logical move. But, ironically, the sale of Altera would benefit AMD. A Way Forward In 2022, AMD acquired Altera’s primary competitor, Xilinx, for $49 billion to have an even more competitive offering for the data center market. This deal has already proven to be successful, and a sale of Altera will throw that company into the typical turmoil that follows most M&A deals. Early this year, Intel made efforts to give the appearance that its manufacturing business is separate from its design and products business. It began reporting its “Product” segment separate from its “Intel Foundry” segment, which made up 26.2% of total revenues in 2023. This move was done to make Intel’s foundry services seem more independent, and perhaps prepare Intel for splitting up the company. That could mean two separate operating entities, or perhaps even the sale of the foundry, but who would buy? Again, Intel is in a pickle. - It missed the mobile market entirely
- It missed the GPU market entirely
- It missed the AI market entirely — no AI-specific semiconductors to speak of that have any commercial traction
- It is years behind in semiconductor manufacturing technology
- And it is gambling on the 18A manufacturing process, for which it has no confirmed volume customer committed (and Broadcom, probably its best opportunity, is not happy with the progress)
The reality is that Intel no longer has the best semiconductors. Its products, while they still bring in about 66% of total revenues, are all replaceable with better products offered by competitors. Even Intel’s most advanced semiconductor — Lunar Lake — is currently being manufactured by TSMC. How ironic is that? Intel would benefit from focus. It needs to choose between being a product company or a manufacturing company. And I would argue that the bigger business opportunity, given Intel’s mistakes, is to become an industry counterbalance to TSMC as a contract semiconductor manufacturer. Sell Altera, sell the 88% of Mobileye (MBLY) that Intel still owns, sell the product division, and raise as much capital as possible to build a semiconductor foundry capable of competing with TSMC. That’s the one thing that the industry needs more of than anything else. And a competitive market with two independent manufacturing giants would promise competitive pricing… And it would guarantee that the exponential pace kicked off by Gordon Moore in the 1970s will continue for years to come, and with it, even more economic and technological progress. | 
Tidak ada komentar:
Posting Komentar