Jim Seto had about ten minutes on the Mining Disrupt Miami 2023 stage, and he spent the first of them on arithmetic. The CEO of ePIC Blockchain put three public charts side by side: the network's seven-day average hashrate, around 390 exahashes per second; the Cambridge Bitcoin Electricity Consumption Index estimate of the power used to mine, about 16.3 gigawatts when the snapshot was taken; and a projection of how soon the network would reach one zettahash. Divide the power by the hashrate, he said, and the fleet in the field is far less efficient than the newest machines suggest.
The keynote, recorded at the conference in Miami on July 25–27, 2023, was a short case for why ASIC makers have to do more than shrink chips, and a description of how ePIC, a company he called "very engineering centric", planned to grow. Bitcoin traded at $29,233 when he spoke. It was $83,079 when this story was written.
A 42 joule fleet
Seto's starting point was the efficiency of every machine already plugged in. Using the Cambridge power figure and the network hashrate, he worked out a nominal midpoint for the whole fleet.
the math just says it's 42 joules per Terra hash
Jim Seto · 3:02
He allowed that the Cambridge numbers may run high. Even on the low side, he said, the midpoint might be 40, or around 38 joules per terahash. Either way, the number told him there was "a pressure point" to build better hardware.
He then pointed to the hashrate chart. The network will reach one zettahash before long, he said, and at a nominal 140 terahashes per miner that works out to more than seven million machines hashing, most of them deployed over two or three years, which is about how long a miner lasts. Seven million machines at a fleet average still near 40 joules per terahash meant, in his words, that the industry needs more innovation and more scale. That was also the message ePIC said it was hearing from customers: whether the work was control boards, firmware or a full system, they wanted it done faster.
What sits inside an ASIC miner
The second slide was aimed at the engineers in the room. People often think a miner is "just a chip and a PCB, a power supply", Seto said. He split it into four quadrants: software and firmware, full system design, chip design, and silicon technology. Process nodes such as seven, five and three nanometers draw the most attention, and he called them interesting and necessary. The harder part, he argued, is elsewhere.
the complexity is probably in the upper right and the upper left in terms of bringing everything together
Jim Seto · 5:28
Chip design is not simple either. Logic design can look like synthesizing a block and building the masks, he said, but reaching the performance and power targets takes work at a lower level.
you really need to start innovating at the design level
Jim Seto · 5:50
The quadrants overlap, and Seto listed where: control protocols between the firmware and the power supply, optimization of the chip within the system, the match between process modeling and what the silicon offers, and telemetry, which he said clients need more of every year. His conclusion was that the companies that understand all four areas are the ones likely to prevail.
Crawl, walk, run
Seto described ePIC's last three years as a crawl, walk, run plan: build each capability in turn, from control boards and firmware to full systems, before taking on the next one. At the time of the talk the company had shipped its first full miner and was about to roll out a second generation.
we are actually have delivered on our first generation Bitcoin miner built here in North America
Jim Seto · 7:06
He said it was public knowledge that the first generation was a partnership with Intel. Anyone who has worked with Intel, he told the room, knows it is large, technical and operationally demanding, and he called meeting its bar "a proud moment" for a small team. The second generation, he said, had a team working on harder problems still.
Helping with other makers' machines
One product surprised some listeners. ePIC had started a fleet enhancement service that optimizes hardware it did not design. Seto said people ask why the company would help a competitor. His answer was that the work is for the client.
our model is really to cooperate with our clients if they're asking us to help them solve a pain Point
Jim Seto · 8:25
If ePIC can add value to a customer's fleet, he said, it will do the work, even when the machines carry another brand.
Why investors listened
Seto said the crawl, walk, run slide was the one that resonated with investors and with people in the capital markets looking to join the company's cap table. In their thesis, he said, much of the risk of investing in a hardware company like ePIC had been taken out because the capabilities were already built.
His last slide set out ePIC's value proposition in three layers. In the short term, clients see gains from fleet and hardware optimization. In the mid term, they see a company that has executed one generation and a second, with a third on its roadmap. In the long term, ePIC wants to work with customers on new density, new form factors and modularity, energy reuse and new geographies, the greenfield work he described as the long-term side of the business.
more Innovation is needed more players are needed and the value proposition of a company like epic is out there
Jim Seto · 10:28
He closed by inviting the audience to booth 25, where ePIC had brought its engineers to meet buyers in person.
Then and now
The keynote is more than three years old. Bitcoin has gone from $29,233 to $83,079 over that span. Seto's numbers are a record of mid-2023: a 390 exahash network, a 16.3 gigawatt power estimate and a fleet midpoint near 40 joules per terahash. His argument, that efficiency gains depend on firmware, power delivery and system design as much as on the next silicon node, is the part to test against today's machines. The full session is linked on this page.