The semiconductor industry has been ablaze with activity this year as the market heats up and established chipmakers face new degrees of competition and a consolidated market. Nowhere is this more evident than in the stark contrasts between Intel, AMD, and Nvidia's latest earnings reports.

Intel ended a string of strong quarters earlier this fall with a downturn in the third quarter that saw the chipmaker's revenues slip 4% and its net incomes plummet nearly 30% year over year. CFO George Davis blamed Intel's tough quarter on “intensifying COVID-related demand impacts, particularly in our data center, enterprise, and government segment.”

These challenges reportedly sent the company's data-centric revenues — which include Intel’s Data Center Group, Internet of Things Group, and Non-volatile Solutions Group — tumbling.

Not long ago, Intel's data center revenues were a growth driver for the company, accounting for more than half the company's quarterly revenues.

Davis blamed COVID-19, but during the company's second-quarter earnings call, Intel revealed its forthcoming 7-nanometer manufacturing process had been delayed. “We are seeing an approximate six-month shift in our 7-nanometer-based CPU product timing relative to prior expectations,” Swan said, according to a transcript. “The yield of our 7-nanometer process, which based on recent data, is now trending approximately 12 months behind our initial target.”

Seven-nanometer chips are now expected to begin shipping sometime in late 2022 or early 2023 — about the time Taiwan Semiconductor Manufacturing Co. (TSMC) is expected to bring 3-nanometer chips to market.

Making matters worse, Intel's upcoming 10-nanometer server chips won't ship until next year. By contrast, AMD's second-generation EPYC processors — based on TSMC's 7-nanometer process — have been available for the better part of a year.

Continued delays have forced Intel to consider the use of external fabrications to meet demand.

By comparison, AMD had a very different quarter, which saw its revenues up 56% to $2.8 billion and net income more than double to $390 million year over year.

Both chipmakers saw strength in their consumer space, but where Intel faltered in the data center, AMD saw growth.

"Much of this growth has come from AMD's success in the PC industry with CPUs and GPUs. The combination of a much-improved product line-up under CEO Lisa Su and Intel's manufacturing misfires have helped AMD win back market share," wrote Geoff Blaber, VP of research at CCS Insight, in blog post.

Key to this growth was the ongoing strength of AMD's EPYC server processors, which offer more than twice the core count of Intel's flagship Xeon Scalable chips, albeit at a lower per-core clock. Announced in August 2019, AMD's EPYC 2 chips offer up to 64 cores and 128 threads. By comparison, Intel's largest chips offer just 28 cores and 56 threads, and this doesn't appear to be changing with the launch of its third-generation Xeon Scalable processors, announced earlier this year, either.

All major cloud providers were quick to adopt AMD's EYPC 2 processors with some already discussing plans to deploy AMD's next-generation Milan chips.

AMD could become an even larger thorn in Intel's side if a proposed $35 billion acquisition of FPGA manufacturer Xilinx moves forward without regulator delay. The acquisition would make AMD the only other large FPGA manufacturer next to Intel.

Beyond EPYC, 2020 has seen AMD take on rival Nvidia in the data center GPU space, a market that has been, until recently, largely uncontested.

The MI100 GPU, announced earlier this month, is the first to go toe-to-toe with Nvidia's new A100 artificial intelligence (AI) accelerator.

AMD claims the GPU is capable of 11.5 teraFLOPs of performance in floating-point 64 tasks. And while analysts don't expect the card to outperform the newly revamped A100 GPUs from Nvidia in AI workloads, they do see the cards as competition in the broader high-performance computing market.

While Nvidia may be best recognized for its gaming roots, today the company is also a big name in the data center, cloud, and for powering some of the most powerful supercomputers in the world.

Beginning last year the company began expanding beyond graphics-based acceleration into new verticals including networking with its $6.9 billion acquisition of Mellanox in 2019 and now CPUs with the planned $40 billion acquisition of British chip designer Arm.

The launch of the company's Ampere GPU architecture in April and the integration of Mellanox paid dividends in the company's latest quarter. Nvidia's revenues for the third quarter of fiscal 2021 topped $4.73 billion, up 57% year over year and 22% sequentially.

Predictably gaming continued to drive the majority of Nvidia's revenues, while data center revenues mounted 1.9 billion, up more than 160% year over year and 8% from the previous quarter.

“Our new Nvidia Ampere architecture gained further adoption by cloud and hyperscale customers and started ramping into vertical industries,” said Nvidia CFO Colette Kress, on the company's Q3 2021 earnings call. “A100 adoption by vertical industries drove strong growth, as we began shipments to server [original equipment manufacturer] partners whose broad enterprise channels reach a large number of end customers.”

Meanwhile, Mellanox saw double-digit sequential growth during the quarter contributing to 13% of the company’s overall revenues.

However, it's not just well-established chipmakers having all the fun. In the CPU space, things are getting more diverse. It's no longer a two-horse race, with more and more chipmakers Arm-based products moving into the data-center market.

Over the past year, several vendors including Ampere, Marvell, and Amazon have brought forward new Arm-based data-center chips, and many of them offer distinct advantages over their older x86 siblings. Chips like Marvell's ThunderX2, Amazon's Graviton, and Ampere's Altra have seen broader adoption over the past 12 months.

Last fall Microsoft announced the availability of ThunderX2 based instances in its Azure public cloud, Meanwhile, Amazon gave its in-house Graviton chip a refresh late this summer, and Oracle announced plans to deploy Ampere's 80 core Altra chips in its cloud data centers.

What all of these chips share in common is Arm's Neoverse microarchitecture, which was designed specifically for use in data center environments. That architecture got a substantial refresh earlier this year. According to Arm, the new chips deliver 40% to 50% higher performance than the previous generation Neoverse N1 while consuming the same amount of power.

And now with Nvidia's proposed acquisition of Arm, the company plans to expand Neoverse development and open its GPU technology for license.