Intel's Altera buy was primarily to bolster its position in the server room. Image source: Intel.
Last year, chip giant Intel (INTC) announced its intention to acquire programmable logic chip specialist, Altera, in a deal valued at about $16.7 billion. The strategic rationale behind this acquisition boiled down to two main themes.
Frist, Intel hoped to give Altera's core stand-alone Field Programmable Gate Array, or FPGA, business a competitive advantage by transitioning the substantial majority of Altera's products away from third-party contract manufacturers and toward its own manufacturing technology.
Secondly, and probably more importantly, Intel wants to be able to integrate FPGA technology directly into its Xeon processors. This, the company argues, should give major customers the ability to configure these processors to accelerate algorithms that are important to customers.
This rationale is solid, but there have been reasons to be skeptical of Intel and Altera's execution. For example, the Intel-built 14-nanometer FPGA, known as Stratix 10, that Altera has been working on appears to have been delayed and will make it to market later than a comparable FPGA from rival Xilinx (XLNX).
At a recent investor conference, Intel CFO Stacy Smith shed some light on why the product was delayed.
Intel puts the blame on Altera
Smith made it clear that the delay of Altera's Stratix 10 was an Altera-specific issue rather than an Intel one. Perhaps more interestingly, though, he indicated that the problem was due to resource allocation within Altera.
"We learned in the diligence [Altera was] having some issues with current generation products," Smith said. "They had to make some resource decisions and so they slowed down the first generation [of products] that's landing on the Intel process technology."
That being said, now that Altera is part of Intel, Smith claims that Intel has worked to "accelerate" the pace at which work is getting done on the first Intel-built FPGA and hopes that the company can "accelerate it a bit more." He continued, "As we get to the next generation, we get back on the [product release] cadence that [Intel] expected."
The stand-alone FPGA is nice, but when will Intel release an integrated part?
It's good to know that Intel is working to iron out the issues that Altera was facing with Stratix 10. In order to realize the full value of this acquisition over time, Intel will need to deliver on its promise of making Altera more competitive -- by using Intel process technology -- than it would have been otherwise.
However, the far more interesting part of this acquisition is the eventual integrated FPGA/Xeon processor product. Intel has said previously that it's sampling parts with a separate FPGA die placed on the same package as a Xeon processor, but a true integrated part -- one that combines the FPGA circuitry with the processor in a single piece of silicon -- is still a while out.
Now that Altera is officially part of Intel, I hope that at Intel's next analyst meeting in Nov. 2016, the company goes into some depth about when it expects such an integrated part to arrive. Given that the acquisition closed in late 2015, and given that it takes several years to design and validate a complex server chip, I wouldn't expect such a product to arrive before 2018, with 2019 or later appearing more likely.
