It’s hard to believe, but the Computer Architecture Podcast is now five years old. We released our very first episode on May 28, 2020, and just like that, we find ourselves at our five-year anniversary, preparing to release our 20th episode. We’ve both been so grateful for the opportunity to do this, and for the support we’ve received from SIGARCH. It’s been a pleasant surprise to see the reach of what began as a humble side project.

We recorded our first episode in person at HPCA’20 in San Diego (featuring the inimitable Kim Hazelwood) with great excitement and trepidation. Could we come up with something people would want to listen to? Would we have any staying power or would the podcast be a flash in the pan? Shortly thereafter the world changed drastically, and all our episodes since have been recorded virtually—a shift that, perhaps counterintuitively, may have helped ensure the podcast’s longevity.

It seems a fitting time to look back and reflect upon our experience in hosting this podcast, which has been a greater success than any of us suspected. As of May 20, 2025, the podcast has been downloaded 53,000 times across 19 episodes. It’s humbling to think that people might be listening to us in all four corners of the world, with our audience members hailing from Estonia, Ghana, Saudi Arabia, Vietnam, and many more countries.

By the numbers:

• Our first episode (Kim Hazelwood on Systems for ML) launched May 28, 2020, and has been downloaded 6.1K times.
• Our most popular episode, featuring Jim Keller of Tenstorrent, has 7.9K downloads.
• We’ve had the privilege of hosting 12 guests who are primarily academics and 7 guests who are primarily from industry, plus one very special episode for the 50th anniversary of SIGARCH featuring a mix of perspectives.
• We’ve featured 6 women and 15 men as guests.
• We’ve featured 3 guests whose primary institutions are outside of North America (Indian Institute of Science, Institute of Computing Technology and Chinese Academy of Sciences, and EPFL).
• In the last 24 months, our listener distribution highlights a truly international community:

• United States: 59.79%
• India: 7.56%
• UK: 5.21%
• Canada: 3.71%
• Germany: 2.62%
• South Korea: 1.96%
• Rest of the World: 19.15%

Over these 19 episodes, we’ve had the privilege of speaking with some of the brightest minds in our field, covering a vast landscape of topics. We’ve journeyed from the intricacies of domain-specific accelerators (Bill Dally) and energy-efficient algorithm-hardware co-design (Vivienne Sze) to the frontiers of quantum computing (Fred Chong) and AI for systems design (Vijay Reddi). We’ve explored privacy-preserving COVID tracing (James Larus), durable security (Todd Austin), the challenges of hyperscale cloud design (Yungang Bao), and the critical need for sustainability in a post-AI world (Carole Jean Wu). Our guests have shared their experiences building full-stack XR systems (Sarita Adve), navigating physically-constrained computing (Brandon Lucia), architecting for Exascale (Gabe Loh), and even the rollercoaster of the hardware startup experience (Karu Sankaralingam, Dan Sorin).

Through these diverse conversations, it’s clear there are some recurring threads of wisdom that come up time and time again:

1. Follow your curiosity. This theme emerged repeatedly, even being codified in the first episode with Kim Hazelwood as “Hazelwood’s Law” as she exhorted listeners to avoid dogpiling on whatever is the hot topic of the moment. Mark Hill also emphasized this as he discussed how he developed his “nose” for good problems, and paraphrased Doug Burger in saying that the “most profound stuff is when you have a contrarian prediction that turns out to be true.”
2. Be open to serendipity. From Dan Sorin’s robotics startup born from a casual lunch with a roboticist, to Brandon Lucia’s satellite collaboration via Twitter (now X) music connections, to Arka Basu’s formative grad school experience with Shan Lu leading to the foundation of his current research program—unexpected relationships often shaped long-term research directions.
3. Collaboration is key. Nearly every guest was open with their gratitude for their collaborators, and acknowledged that their success would have not been possible without mentors and colleagues. In particular, Mark Hill says, “Be humble in the credit because the goal really is to have renewed collaboration…Nobody is going to do a recurring collaboration carrying the other person.”  Jim Larus’ COVID tracking technology is a case in point–a global team of 30 cross-disciplinary experts, from technologists to epidemiologists, came together entirely over Zoom and Slack to build their widely-used public health tool, leading directly to the next theme.
4. Communication and non-technical skills are incredibly important. The idea of building a global team over virtual channels at a time when this was not the norm required excellent communication and interpersonal skills. No whiteboards and no in-person meetings, and yet the result was something beyond a publication–it was a widely deployed and important tool impacting global public health.  At the same time, Vivienne Sze says that she has run a workshop at MIT for leadership skills because “…building those types of skills—managing feedback, conflicts—these non-technical skills are really important at every point of your journey.” She adds that these are skills she uses every day, and that being adept at this frees her mind for technical work. And more from Mark Hill, the Yoda of our field, “I can’t emphasize enough how important written and oral communication are…Our job is not just to discover things but to communicate them…Even criticism you think is wrong—ask how did I miscommunicate that led to this?”
5. Interdisciplinary work is required for broad impact. In a post-Moore’s Law world, crossing boundaries is essential, particularly in newer fields like AI, quantum computing, and AR/VR/XR where interfaces and abstractions have not yet been hardened. Cross-layer optimization is important for overall system success, and a prime example is Sarita Adve’s description of her XR work, “XR touches so many domains — graphics, computer vision, robotics, haptics, optics… You have to involve people from all of these areas. It’s super challenging, but super fun…True co-design, cross-layer optimization…you cannot solve this by just working in hardware or compilers or runtime—it all has to come together.”
6. The journey from research to product is much more than a cool idea. This came up over and over again, and hinges on bringing an idea beyond ideation and proof of concept to making it a reality.  Bill Dally puts it crisply, “A big reason tech fails to reach a product is that it only gets far enough for a paper, not a production GPU…there’s a big difference between something done enough to publish a paper and something done enough to bet a GPU on.”  The difference between cool tech and true customer value was also succinctly described by Dan Sorin, “We had tech where people would come in, see a demo and go, ‘I want that.’ Then they’d see the product and say, ‘I don’t want that.’”  But true impact is so worth it— as Dan puts it, “Now that I’ve seen the process of taking technology into the real world, it’s a little tougher to get excited about research unless I can see where it might go.”  Beyond the importance of true technical maturity, there is also the importance of true customer value (and communicating that value).  Karu Sankaralingam told us, “The business part is absolutely the hardest.  Getting customers to believe in a long-term value proposition…took a lot of conversations.”
7. Software is a key part of successful hardware.  As several of our guests forge ahead in new areas and building novel systems, the software stack has critical importance, with several sub-themes:
   1. Software as the enabler (tooling): A lot of hardware development is gated by mature software tooling.  Fred Chong says of quantum computing, “you need debugging tools, visualization, and performance monitoring — even though we’re far from scaling, software matters today.”  Brandon Lucia similarly emphasized the need for a robust tooling stack in his work in physically constrained computing, starting from the compiler. “The compiler needs to exist…that’s step zero. If you don’t have that, the thing won’t get out of the starting gate.”
   2. Software as the peer of hardware in designing the optimal system: Hardware/Software co-design can lead to better optimizations on both ends of the spectrum.  As Sarita Adve put it, “We’re designing a whole stack, starting with how we represent the code…so it maps to diverse hardware.”
   3. Software as the ultimate consumer touchpoint: Complete software support for a robust user experience is table stakes for customer buy-in.  As Karu Sankaralingam told us, any hardware technology can fail commercially if a very high-level of user experience is not met, leading to a huge fraction of his hardware startup’s resources being devoted to writing “a boatload of software.”

Hosting this podcast has been an incredible learning journey for us. It has deepened our appreciation for the breadth and depth of computer architecture, and for the passion and ingenuity of the people driving it forward. We’ve been consistently inspired by our guests’ willingness to share not just their technical achievements but also their personal journeys, their triumphs, and their lessons learned.

We are immensely grateful to SIGARCH for its unwavering support of this initiative from the very beginning, particularly from our primary liaison to the executive committee, Boris Grot. We extend our gratitude to then-SIGARCH chair Babak Falsafi, who originally supported our pitch, and to Partha Ranganathan for his early connections and constant encouragement. We also extend our heartfelt thanks to all our distinguished guests for generously sharing their time and wisdom, and to our listeners around the world for tuning in.

We’re excited to continue bringing you closer to the cutting edge in computer architecture and the remarkable people behind it. Be on the lookout for our upcoming 20th episode featuring Ricardo Bianchini. Here’s to the next five years!

About the Authors:

Lisa Hsu is semi-retired and divides her time between working part-time in Reality Labs Research at Meta on optics and display technologies for AR, playing tennis, and shuttling her children to extra-curricular activities.

Suvinay Subramanian works on the architecture and codesign for Google’s ML supercomputers, Tensor Processing Units (TPUs).

Disclaimer: These posts are written by individual contributors to share their thoughts on the Computer Architecture Today blog for the benefit of the community. Any views or opinions represented in this blog are personal, belong solely to the blog author and do not represent those of ACM SIGARCH or its parent organization, ACM.