It’s all about the energy, baby!
The Fusion Report has three big conferences happening this week to talk about: NVIDIA GTC in San Jose, EV Charging Expo in Las Vegas and the FIA Policy Conference in Washington DC, in as little as 1,500 words (total, not each!). But to start out, let’s talk some units and cross-measurements that are relevant:
- Watts (W) are a unit of power produced, consumed, or stored per unit of time. Specifically, watts are a measure of joules per second (J/sec).
- A kilowatt (kW) is a thousand (103) watts, while a megawatt (MW) is a million (106) watts, and a gigawatt (GW) is a billion (109) watts. A kilowatt-hour (kWh) is the amount of energy used when burning one kilowatt for one hour (3.6 MJ).
- A food calorie is actually a kilocalorie (kcal), or the amount of energy needed to raise 1 kg of water by 1°C, which equals 4,184 joules. The recommended daily allowance (RDA) is 2,000-2,500 kcals/day, or 97 W to 121 W. In comparison, a competitive marathon runner will take~3 hrs to complete the race, and consume 3,000 kcals (1,162 W) while running the race (12.5 MJ, or 3.5 kWh).
- The NVIDIA Rubin GPU consumes ~2.3 kW/chip, or 150 kW for a server rack.
- A Chevy Bolt EV going 60 MPH on a flat road consumes ~15 kW of power, or the same as ~6.5 Rubin GPUs. A GPU rack could power 10 Chevy Bolts at 60 MPH.
NVIDIA GTC (San Jose, CA)
NVIDIA’s GPU Technology Conference GTC) is their flagship AI event, where CEO Jensen Huang kicks things off with a high-profile keynote on new chips, platforms, and software. The show brings together developers, researchers, startups, and executives for sessions on topics like physical and agentic AI, robotics, data centers, and autonomous machines. While GTC originally was concerned with computer graphics, it has now focused on AI problems like LLM solutions and deployments.
While NVIDIA has become ubiquitous with artificial intelligence, they are not the only solution for AI chips these days. NVIDIA’s competition has moved from AMD (who initially purchased the graphics CPU business from ATI) to Google, Amazon and Huawei (note that AMD still has a small chunk of the AI chip business). In fact, NVIDIA competitors have over 30% of the total business as the market shifts from AI training (which NVIDIA chips are optimized for) to inference (which some competitors are potentially better at). From a purely economic standpoint, some of NVIDIA’s largest customers, including Google and Amazon, probably want to keep more of their money in their own pocket as opposed to putting it in NVIDIA’s. It is a repeat of what has happened in the server space and in the NIC space, where Google, Amazon and Microsoft have built their own chips.
Source: Epoch AI
From a power consumption standpoint, efficiency is the key goal for NVIDIA and the other chip manufacturers. The measure of output for AI and similar workloads is called the token. From a comparison perspective, NVIDIA Rubin processes 10X per watt that the number of tokens that an NVIDIA Blackwell (NVIDIA Rubin’s predecessor) does. A 1GW “factory” of NVIDIA Rubin chips (having roughly 100,000 to 150,000 Ruben GPUs) can process roughly 700M tokens/sec, while consuming 24 gigawatt hours (GWh) of energy in a single day. Turning tokens into queries is not straightforward; a small query typically takes 100-200 tokens (per ChatGPT, of course), while the maximum size of a GPT 4o query is 128,000. By comparison, that would be enough electricity to fully charge 370,000 Chevy Bolts electric vehicles (EVs), since each Bolt stores 65 kWh, and has a range of approximately 250 miles each. Alternatively, that would be enough power for 800,000 average US homes, with each consuming 30 kWh per day.
Zooming out a bit for context, it is interesting to look at the number of AI prompts utilized daily in the US. ChatGPT, which makes up 48% of all AI traffic, has a volume of 330 million prompts daily in the US (slightly less than one person per day), and 2.5 billion per day globally. While this sounds compelling, Google itself handles approximately 16.4 billion searches per day, 6.5 times the amount of AI traffic globally. However, by 2030 that rate is expected to be roughly even between standard searches and AI queries. That is why Dell’Oro found that data center CAPEX surged 57% in 2025, and is expected to top $1 trillion in 2026. Not a bad business to be in…
EV Sales – A Rollercoaster in the US, But DC Fast Chargers Going Hot
In spite of the end of federal tax subsidies by the Trump administration, electric vehicle (EV) sales in the US reached their second highest level ever, just 4% behind 2024. 2025 saw sales rates between 1.5 and 1.6 million EVs in the US, reaching 438,000 units in Q3 2025 before the numbers plummeted in Q4 2025 due to the end of the federal tax incentives. That is in contrast to the traffic in China and Europe, where EV sales numbers grew year-over-year (YoY) by 30% and 17%, respectively. Despite the slight decline in US sales, all is not bad by far, as the price difference between EVs and non-EV vehicles shrinks rapidly. In early 2026, the difference between EVs and internal combustion engine (ICE) vehicles dropped to a record low of $6,532, largely spurred by competition (watch out Tesla), and falling battery prices. More dramatically, used EVs have dropped to a difference of only $1,376 in January of 2026. With lower fuel and maintenance costs for EVs, they could actually be cheaper than ICE vehicles from a total cost of ownership.
On the EV charger side, particularly for fast DC chargers, the picture is much rosier. According to Fortune Business Insights, the North American DC charger market is expected to grow at a CAGR of 24.0% between 2025 and 2032, growing from slightly over $1 billion to over $4.5 billion during that timeframe. With EV charging time between 15 and 30 minutes for high-power DC fast chargers, the difference in time between charging an EV versus filling one’s gasoline tank has dropped significantly and is now at effective parity. The current leader in DC fast charging stations in the US is the Tesla SuperCharger network, followed by Electrify America and EVgo. Additionally, European heavyweight Mercedes-Benz is looking to stand up a DC fast charger network in the US, and the replacement of low-power public AC chargers is continuing. Which makes for good times at the EV Charging Summit and Expo…
The 2026 Fusion Industry Association (FIA) Policy Conference
While our article can’t provide a live perspective for the Fusion Industry Association (FIA) Policy Conference since it starts Wednesday morning, we can still provide some insights as far as what is expected to be discussed there. A large part of the discussion tracks will focus on the priorities in several non-US markets, including the UK, Germany and Japan. With the backdrop of the U.S.-Iran conflict, securing future energy needs in all three of these markets is a primary concern and fusion energy, while not yet “next year’s power source”, it is not very far off.
From the UK perspective, there will be a panel discussion titled “Exploring the United Kingdom’s Fusion Priorities”, which includes Neil Cosgrove (Strategic Projects Director of the UK Office Investment) and David Sanford (Deputy Director of Fusion Energy at the UK Department of Energy Security and Net Zero). The UK government’s recent announcement of £2.5 billion ($3.3B US) expenditure on fusion energy will certainly figure into this conversation. A similar session titled “Japan’s Fusion Priorities” includes Akira Tsuneto, the Deputy Director General for Science, Technology and Innovation Policy at the Japan Cabinet Office, while similar panels covered the efforts in Germany and the rest of the EU. Other sessions covered the granting of construction permits for Helion at Malaga, WA, and the Type One Energy construction plans at the TVA Bull Run site in Anderson County, Tennessee.
Why Does All This Matter?
What we put our money into, and how we utilize our energy are the real issues. We live on a planet that does not have enough available resources in the short term for all of its numerous needs. Luckily, there are people willing to take a chance to develop new solutions that can provide more resources when they are needed, whether that is new solutions, new chips, ways to power vehicles, or the development of new energy sources such as fusion energy in its various forms and incarnations. By utilizing new technologies and assiduously managing risks, mankind has been able to solve a lot of problems. There’s no reason this needs to be any different.
