The IT industry absolutely loves new products and technical achievements but like any obsession, it occasionally turns pathological. A good example is the assumed superiority of emerging technologies over well-established products and services. As a result, leading vendors, like IBM, Cisco and Oracle are regularly depicted as circling the drain due to ascendant plucky upstarts.
Oft times, these breathless claims carry more than a hint of competitive wishful thinking, like the past quarter century of claims that IBM’s mainframe business is on the verge of collapse. Despite the melodrama, these vendors (with a few tweaks and adjustments) continue to motor along driving tens of billions of dollars in annual revenues and profits.
This has also been in-play around Intel whose critics cite the company’s various challenges and the rise of new silicon technologies as proof of an imminent collapse. Instead, those vaunted innovations have mostly faltered or fallen flat (remember when Arm server chips were going to own 20 percent of the market by 2020?). Meanwhile, Intel continues to deliver superior solutions and solid-to-stellar financial results.
That doesn’t mean the company has done everything right or is idling along like an overly-entitled cruise ship. Former CEO Andy Grove’s classic line, “Only the paranoid survive,” is baked into Intel’s DNA. So, it was hardly surprising that the company used its recent Architecture Day event to discuss its plans and strategies for the coming decade of “data-intensive” computing.
Six strategic pillars
During the event, Intel detailed half a dozen “strategic pillars” it will rely on to continue its technical leadership and capture business in emerging “data intensive” markets and use cases, including next gen PCs and consumer devices, high speed networks, ubiquitous artificial intelligence, specialized cloud computing and autonomous vehicles. The six pillars are:
- Process: Emphasizing Intel’s manufacturing innovations, including demonstrating the new “Sunny Cove” 10nm CPU architecture and “Foveros” its “industry first” 3D packaging technology which is designed to allow products to be broken up into smaller “chiplets” that can be organized and connected to enhance specific features and performance.
- Architecture: Rather than focusing largely on its traditional IA scalar architecture, Intel envisions a future where diverse scalar, vector, matrix and spatial architectures are deployed in CPU, GPU, accelerator and FPGA sockets as needed. In turn, these architectures will be uniformly supported via a unified scalable software stack and integrated with Intel’s advanced packaging technologies.
- Memory: Given the growing demand for high-capacity, high-speed storage, Intel believes its Optane technology will enable it to uniquely fill gaps in the memory hierarchy and provide enhanced bandwidth closer to the silicon die.
- Interconnect: The company plans to offer a complete line of scalable offerings from silicon-level package and die interconnects to 5G infrastructure solutions.
- Security: Intel believes it has the necessary components to build a “better together” strategy that improves end-to-end security.
- Software: According to Intel, for every order of magnitude performance improvement enabled by hardware, software can enable two orders of magnitude improvements. Among other efforts, the company is planning a common set of tools for developers, including a “One API” project designed to simplify programming across diverse CPU, GPOU, FPGA, AI and other accelerators.
Unsurprisingly, these strategic pillars emphasize Intel’s longstanding leadership in semiconductor manufacturing and investments in new and emerging technologies. This former point is worth emphasizing, especially since most of Intel’s core competitors have replaced their own silicon production with third parties, like TMSC and GlobalFoundries.
That makes financial sense, at least in the short term. After all, building and supporting a chip fab is always complex and costly, especially as semiconductors become ever smaller and older 200mm and below wafers are replaced by current state-of-the-art 300mm wafers. Given the ongoing funding requirements, signing-up with third party fabs seems eminently prudent.
But it also opens the door to unexpected challenges if thosepartners succumb to problems of their own or are rocked by events beyond theircontrol, like natural disasters and global trade disputes. Interestingly, a few days after Architecture Day, a blog post by Dr. Ann Kelleher SVP and GM of Intel’s Manufacturing and Operations organization, detailed how the company is investing inexpanding its 14nm manufacturing capacity to increase supply.
Kelleher also noted that Intel is in early planning stages to expand its global manufacturing facilities in Oregon, Ireland and Israel, and will also continue its two-decades-long strategic use of foundry partners for specific technologies. With what has been estimated to be a $300B total addressable market for silicon solutions, Intel is doing all it can to ensure that it can and will reliably meet future customers’ demands.
So, the strategic vision and practical steps Intel discussed during its Architecture Day proceedings show the company well-positioned to take full advantage of current and future opportunities, right? Mostly, yes, excepting some lines that need further definition and/or coloring-in.
For example, though its security strategy is intriguing, there could be devils lurking in the details. Security is a notably tough nut to crack for IT vendors and more strategic security initiatives fail than succeed. Those include Intel’s 2010 $7.7B acquisition of McAfee which never produced hoped-for synergies, leading to the company selling a majority stake in its McAfee assets TPG for $3.1B in 2016. How Intel delivers on its “better together” security vision should be interesting but it will also be closely scrutinized.
Another point: Though Intel’s Optane memory technology has been promising, computer memory is a notoriously volatile business. In fact, some knowledgeable analysts (including Jim Handy of Objective Analysis), are anticipating a collapse of DRAM prices in 2019. Should that occur, it could result in significant losses for Optane (which is priced to be competitive with DRAM). Intel’s steady profitability would help minimize the pain but no one wants to incur significant losses.
Those points aside, Intel’s Architecture Day provided notable insights into the company’s ongoing efforts and upcoming plans that should cancel out rote disparagement from critics and competitors. Of particular interest were the demos of the upcoming Foveros 3D packaging technology and Sunny Cove CPU architecture. If those innovations deliver what Intel suggests, the future will continue to be bright for the company’s Core and Xeon portfolios.
Along with Foveros, the One API software project should be of particular interest to Intel’s OEM and other customers. People tend to forget that IT and consumer electronics vendors whose solutions depend on third party chips and other components are always in the hunt for ways to lower costs and improve efficiencies. Using Foveros to effectively mix/match chiplet blocks should pave the way to enhanced new silicon solutions, as will reducing the complexity of developing software for diverse computing engines.
Overall, Intel’s Architecture Day clearly highlighted the ways in which the company is anticipating and preparing for the future. But it also revealed that far from passively waiting for that time to arrive, Intel is actively working to make the future it envisions into a reality.