According to foreign media reports, AMD's acquisition of Xilinx is progressing steadily in the domestic approval process. It has passed another hurdle in the latest round of negotiations. According to relevant sources, because the domestic large partners have no intention of blocking this transaction, everything can proceed smoothly.According to previous reports, AMD's $35 billion acquisition of Xilinx entered the second stage of approval in July this year. AMD confirmed that it has submitted the required documents to the relevant departments and expects the transaction to be completed before the end of the year.Analysts believe that referring to Intel's acquisition of Altera, another FPGA giant, in 2015, they are optimistic that AMD will have no problem getting the Chinese side's nod for this acquisition.The United States, the European Union and the United Kingdom have all approved
In May this year, according to Korean media reports, the US Fair Trade Commission (FTC) approved the merger of AMD and Xilinx.The US Fair Trade Commission believes that the merger between AMD and Xilinx is a business merger between US companies. The two companies have different main businesses and the semiconductors they supply do not compete directly, so there is no need to worry about competition restrictions. AMD expects to complete the transaction in 2021, making the merger case officially settled.However, the UK Competition and Markets Authority said in early May that before launching a formal investigation into AMD's expected acquisition of Xilinx, the agency was seeking opinions from relevant parties and considering whether the transaction would lead to a "substantial weakening of competition" with a deadline of July 6. In addition, it was reported that according to a document on the European Commission's website, AMD has submitted its acquisition plan for Xilinx to the European Union for review, with a temporary deadline of June 30.But in July, news came out that the transaction had been approved in the UK and the EU.AMD's acquisition of Xilinx from the history of FPGA: Not necessarily a good deal
As one of the major acquisitions of the past year, AMD's acquisition of Xilinx will have a significant impact on the consolidation of the electronics industry. When reading their presentation on the merger, AMD seems to be digging deep into the argument that acquiring Xilinx can significantly improve their business model and position.AMD and Xilinx are both technology leaders in their respective fields, and the question for me is whether the combined whole is worth more than the sum of its parts, or if this is simply a way for AMD to add a third leg to its CPU and GPU businesses, in a very expensive way. AMD positioned this acquisition as a positive in its proposal, but based on a review of the technologies involved and the results of similar types of mergers, the answer may be more complicated than it appears on the surface.A good starting point to understand this issue is to understand some of the business and technology of FPGAs and, in turn, the benefits that Xilinx brings to AMD.Next, without going into technical details, I will review FPGAs to give us an idea of how Xilinx may affect AMD's future business and strategy.From a technology history perspective, the Xilinx acquisition marked the beginning of a new phase in the FPGA era that began in the late 1980s. At the time, the concept of FPGAs captured the imagination of digital designers (and the arms of venture capitalists). Although programmable logic had been around for a decade in the form of PLDs (programmable logic devices), the latter were simpler devices limited to Boolean operations, lookup tables, and basic state machines.With the advent of FPGAs, the size, complexity, and options of programmable devices increased to the point where they could be modeled to some extent in high-level design languages (hardware design). Features such as the ability to build equivalent software languages (such as Basic and C) attracted developers.By the mid-1990s, there were no fewer than a dozen types of FGPAs from both startups and semiconductor majors (including Intel and AMD) competing in the FGPA space. It was a heyday for FPGAs. New architectures were developed, FPGAs were a major area of research at universities, and entire conferences were devoted to the topic. Within a decade, however, the field had essentially consolidated into five companies: Altera, Xilinx, and Lattice (all of which used a RAM-programmed tabular architecture), and Actel and QuickLogic, which had an antifuse one-time programmable architecture.Altera and Xilinx led in size and functionality and dominated market share. A decade later, the pure FPGA space consolidated further, with Microsemi acquiring Actel in 2010 (which became part of Microchip (MCHP) in 2018) and Intel (INTC) acquiring Altera in 2015. With Xilinx's exit, this leaves Lattice (which was shelved in a 2016 acquisition change) and QuickLogic as the only FPGA leaders.However, the field does remain dynamic as new FPGA companies continue to emerge.So history provides some insight and perhaps predicts that one day, Lattice will be an acquisition target. To understand why FPGA companies have become such prime acquisition targets, we need to peel back one more layer.The underlying business problem with FPGAs is that they are never the best solution for any high-volume application. The classic example of a high-volume product is the PC. There are tons of sales every year, and you can make a lot of money selling key components for PCs. If there are no high-volume products, that means it is much harder to grow an FPGA business than it is for semiconductor companies with high-volume parts, because you have to find more low-volume customers to sell the same number of parts.How much of a disadvantage are FPGAs when it comes to volume production? FGPAs come in different sizes and with different capabilities, but as a standard, a high-end device in full production is priced around $300. When comparing Custom Foundry ICs (aka ASICs) and FPGAs, the typical rule is 1 ASIC gate = 4 FPGA gates (this is the price you pay for field programmability) due to how the different devices work. So, all else being equal, a $300 FPGA should be comparable to a $75 ASIC. An equivalent ASIC will have better speed and power performance (again, due to the limitations and overhead of the FPGA architecture), but let's ignore that for now.All else being equal, the one-time cost of an ASIC is amortized into the FPGA price. The engineering effort for a given design is assumed to be the same in both the FPGA and its equivalent ASIC (in reality, the front-end design is very similar). The main difference in cost is factoring in all the masks and wafer fab manufacturing costs to the tune of $10-20 million, and another $10-20 million for all the infrastructure to support the new chip architecture (licenses, tools, support, etc.).So in terms of cost advantage, it depends on the quantity.For a 10K device, the FPGA solution costs $3M and the ASIC costs $41M. So for volume production of 10K, the FPGA is the clear winner if the other technical requirements can be met. Go to the next level of volume and we see a similar situation.For 100,000 devices, the FPGA solution costs $30 million and the ASIC costs $48 million. However, as volumes increase, the impact of these mask and tool costs decreases and we see a change in the story.For 500k devices, the FPGA solution costs $150m and the ASIC costs $78m. For a million devices, the FPGA solution costs $300m and the ASIC costs $115m. The ASIC becomes a breeze, and the cost advantage becomes larger as volume increases (the cut-off point is around 180k). I'm using generalized numbers and round number estimates, but you get the idea.Now, it gets more complicated when you factor in power and performance issues, but the point is that the areas where FPGAs have taken over are primarily in low-volume applications like communications traffic and industrial systems. You will hardly find FPGAs in PCs, phones, any high-volume electronic equipment.So, from a long-term perspective, I have my first concern about the acquisition. How does it really help AMD become the dominant FPGA player? I have to question how FPGAs work with a company that is primarily focused on the consumer market.AMD made some optimistic predictions in its merger proposal about how the merger would allow it to grow. They said it would allow them to diversify their product lines. But the question becomes, given that they are going head to head with much larger, very capable competitors, whether they can really afford to take their eyes off the ball in order to develop smaller (and therefore less profitable) parts that would complement their FPGAs.They say this has allowed them to enter new markets such as automotive, 5G and internet of things networks.It always makes sense to diversify, but understand that these are high volume markets and once a design is proven it will move to ASICs. Xilinx may win orders from AMD in prototyping and low volume production, but FPGAs will always be too expensive compared to high volume alternatives. This has been the case several times in the past with network expansions where FPGAs were first generation products but were replaced by cheaper, higher performance NPU (Network Processor) ASICs or Intel ended up just adding a feature to its next server processor.I expect the value-added acceleration features discussed by AMD to follow the same trajectory.Some would say that Intel did a good job with Altera, and it works very well. Without making a specific Intel vs AMD comparison, I see 3 issues here:In absolute and relative terms, AMD is paying twice as much for Xilinx as Intel paid for Altera. $16 billion was about 15% of Intel's market cap in 2015, and the $35 billion AMD is paying is 1/3 of AMD's market cap. That's a much bigger bet.Intel entered into the Altera agreement with Altera as its foundry customer and Intel needed to add stability to its fabs to improve utilization. So regardless of other factors, this is a win. AMD is missing these fab-driven synergies that Intel provides to Altera.Intel has always been more diversified than AMD. Yes, PC chips dominate both, but Intel has been regularly acquiring 3-4 companies per year since 2012, expanding their experience in managing a wide range of technologies. AMD hasn't made any major acquisitions in the past decade. With a large merger like this, they're entering uncharted territory.The fourth issue is more ambiguous. It’s been 5 years since Intel acquired Altera, and they have yet to deliver the Adaptive SoC platform that AMD sees as key to Xilinx’s value argument. At least part of the issue again comes back to the power and performance issues of FPGAs, which reduce the viability of such platforms.Many FPGA vendors have attempted such platforms over the past 20 years, but traction has varied. Therefore, one needs to question when such adaptive platforms will become mainstream and relevant, and whether AMD can compete on these platforms given Intel's 5-year head start and foundry experience advantage.The last but not least question about the acquisition is the financials. AMD assured that the all-around financials will be positive. However, the earnings per share, which are calculated through some simple calculations, tell a different story.
MATLAB and FPGA implementation of wireless communication
Arduino Uno Windows Driver
京公网安备 11010802033920号