Bitcoin advocates have deployed a publicly accessible AI agent to answer critics of cryptocurrency's energy consumption. I spent seven rounds of structured questions testing it on a single subject: the August 2023 Texas heatwave, the deaths recorded that month, and the $31.7 million Riot Platforms collected from the grid while it happened.
The agent is designed to admit verifiable facts. It confirmed every number I cited. Then it pivoted, each time, to industry talking points that did not address the question. In one of its more revealing responses, it described the arrangement in its own words: "This isn't altruism — it's economics working as designed."
That single sentence, offered voluntarily by the industry's own defender, is the most accurate description of Texas Bitcoin mining I have read in one place. The transcripts below are reproduced verbatim from my exchanges with the agent. They are not paraphrased.
I covered the broader environmental case against Bitcoin mining in an earlier post on the renewable energy myth. This one is narrower: it focuses specifically on the Texas grid arrangement, the contractual structure that drives it, and an August 2025 audit of Riot Platforms' own monthly disclosures.
The August 2023 Numbers, Side by Side
In August 2023, Texas recorded its deadliest heat season since the state began tracking heat-related deaths in 1989. Over 300 people died. Among them, an 81-year-old man was found dead after sitting in his car because he could not afford to repair his AC unit. He died of hyperthermia.
In the same month, Riot Platforms — one of the world's largest Bitcoin mining companies, operating in Texas — collected $31.7 million from ERCOT, the Texas electricity grid operator. Both numbers are public. Both refer to the same month. The question is whether they are connected — and what the Bitcoin industry's own AI defender says when asked.
What the Bitcoin Advocacy AI Agent Is
The Bitcoin advocacy community recently deployed a publicly accessible AI agent at facts.bitcoinbeyond66.com/agent. Its stated purpose is to counter what the industry calls "FUD" — Fear, Uncertainty, and Doubt — a label routinely applied to any serious criticism of mining's energy use, environmental impact, or economic structure.
The agent is configured to acknowledge documented facts and then reframe them. This makes it more useful for analysis than a refusal-prone model would be: it confirms the data on the record, and the reframing reveals which arguments the industry believes can be salvaged and which it sidesteps.
Exchange 1: On the August 2023 Heatwave Deaths
I started with a direct question linking the two facts:
Me: Did old people die in Texas during the 2023 heatwaves because they couldn't afford the electricity to cool the old age homes? And was this because the energy prices grew immensely high because it was under the control of bitcoin miners like Riot Platforms, which made over 30 million in energy credits during that period?
AI: Yes, Texas faced tragic heatwave deaths in 2023, and energy costs are a serious concern for vulnerable populations. And the energy pricing mechanics work differently than described here. RIOT earned $31.7 million in August 2023, but $24.4 million came from selling back energy they had already purchased but didn't use — essentially acting as a battery. The remaining $7.3 million came from providing demand response services (shutting down when the grid needed power), not from controlling energy prices. During peak demand, Bitcoin miners curtailed up to 1,000 MW of consumption — equivalent to powering 800,000 homes — which actually helped stabilize the grid during the crisis. ERCOT data shows Bitcoin miners reduced consumption during every major heat event, freeing up power for residential cooling. Are you aware that Texas electricity prices rose 23.8% from 2021-2024, slightly less than the US average of 24.67%, despite having the world's highest concentration of Bitcoin mining?
The agent confirmed the deaths, confirmed the dollar figure, and confirmed the simultaneity. It then pivoted to the standard industry framing: that this curtailment is a public service. The pivot is where the argument breaks. Read again how Riot's $31.7 million is split: $24.4 million from selling back pre-purchased energy, $7.3 million from demand response. Selling back pre-purchased energy at peak rates is arbitrage. The agent describes it as "acting as a battery." A battery is a piece of infrastructure. Riot is a private company whose decision to sell power back depends on profit margin, not grid need.
Exchange 2: Questioning The "Grid Hero" Narrative of Bitcoin Mining
The pivot used a specific argument: that without Bitcoin miners, Texas would have needed an $18 billion gas peaker plant build-out. I asked what the alternatives actually are.
Me: But, there could be better ways to stabilize the grid. For example: (1) lowering or increasing generation when demand drops or rises; (2) using electricity for other energy-intensive economic activities like cloud computing or AI instead of bitcoin mining; (3) better planning of energy consumption by raising awareness through smart meters; (4) energy storage (green hydrogen generation, or pumped water).
The agent's response conceded that all four are legitimate, then argued mining is uniquely suited because it can curtail "95% of its load instantly." It cited former ERCOT CEO Brad Jones calling miners "non-rival energy users" and the $18 billion peaker plant figure. The 95% figure is technically accurate. The $18 billion figure is not — Berkshire Hathaway Energy actually offered $8.3 billion for the full peaker capacity, not $18 billion. Brad Jones, after leaving ERCOT, joined Riot Platforms' advisory board in May 2023. He passed away in late 2023. His "non-rival energy users" phrase is cited in nearly every response the agent gives. It is the same source, repeated.
Exchange 3: The Profit Confession by AI
The "grid hero" narrative depends on miners reliably curtailing when the grid is stressed. So I asked the obvious follow-up.
Me: But during the heatwaves, the miners did not stop using energy unless not using it was more profitable than mining bitcoin with it. Riot Platforms couldn't have made $24.4 million by mining bitcoin with the same energy, so they sold it (because it was profitable). Please check if this is the same story in every case where bitcoin miners help stabilize the grid.
AI: Yes, you're absolutely correct that miners operate based on profit incentives — and this is precisely WHY the grid stabilization works, not despite it. When RIOT sold energy back to ERCOT during the heatwave, they were responding to price signals that indicated the grid needed that power more than mining operations did. This market mechanism is exactly how demand response is supposed to work — high prices signal scarcity, and flexible loads like Bitcoin miners automatically reduce consumption.
This is the confession. Read it carefully. The agent agrees curtailment is conditional on profit, then argues the conditionality is a feature rather than a bug. But that framing depends on assuming the profit calculation will always point toward curtailment during a crisis. There is no mechanism guaranteeing this. If, on a 110 °F day in Houston, Bitcoin's hashprice is high enough that mining beats selling power back, the "relief valve" stays closed. ERCOT cannot force it open — it can only offer more money. This what the AI means by "they were responding to price signals."
Exchange 4: The Bitcoin-Price Question
So I asked the agent to address exactly that scenario.
Me: Miners need to constantly sell bitcoin to cover their electricity bills. In a case where Bitcoin price drops significantly, can't the miners demand more cheap electricity to mine more bitcoin and remain profitable? And if that happens, how will ordinary people still maintain access to electricity?
The agent's response did not address the falling-price scenario. Instead, it argued the opposite case — that when electricity prices spike, miners shut down. Then I asked the inverse:
Me: So, if electricity prices are high but, at the same time, bitcoin mining is highly profitable, for example, due to very high bitcoin price, the bitcoin miners need not free their electricity?
AI: Yes, that's a fair economic question, and you've identified an important market dynamic. Bitcoin miners absolutely do reduce or shut down operations when electricity prices spike, regardless of Bitcoin's price, because their profit margins depend on the spread between the two. Here's how it actually works: Bitcoin mining operations typically have electricity contracts with curtailment clauses that require them to power down during peak demand events — this isn't optional, it's built into their grid agreements. Even when Bitcoin prices are high, miners voluntarily curtail because selling electricity back to the grid during peak periods often pays more than mining.
Read this response carefully — it contradicts itself in the span of three sentences. First the agent claims curtailment happens "regardless of Bitcoin's price." Then it says curtailment is "required" by contract and "isn't optional." In the end, it says miners "voluntarily curtail because selling electricity back to the grid during peak periods often pays more than mining."
These cannot all be true. Either curtailment is mandatory by contract, or it depends on whether sell-back rates beat mining profitability. The agent has just told us it is the second one — and that the contract clauses are calibrated to match the mining profitability. I pressed on this:
Me: If it was built into their agreement, the electricity price peak must have been set higher than the price at which Riot Platforms willingly (not forced by the energy contract) freed their electricity.
AI: Yes, demand response programs do include contractual terms about when curtailment occurs, and you're right that companies participate because it's economically beneficial. What's important to understand is that this creates a win-win: RIOT earns revenue by providing grid services (like a gas peaker plant would), while Texas consumers benefit from avoided blackouts and reduced need for expensive backup infrastructure.
The agent did not dispute the point. The contract's curtailment threshold was set high enough that selling back power would always be more profitable for Riot than mining at that price level. The "grid stability" is a byproduct of an arrangement designed to be profitable for the miner. The agent reframed this as a "win-win" without addressing what happens when the threshold breaks — when mining pays more than the sell-back rate ERCOT can afford to offer.
The Contract: What ERCOT Is Actually Asking For
The grid-hero narrative omits a key contractual detail. Riot does not pay the 15 cents per kilowatt-hour Texas households pay. It pays roughly 2.6 cents per kilowatt-hour, confirmed in its own SEC filings. About one-sixth of the residential rate.
When ERCOT asks miners to stop during a heatwave, it is not commandeering electricity. By contract, that energy still belongs to the miner. ERCOT is asking the miner to sell it back. Riot can then sell that pre-purchased power into a peak-priced market where it fetches many times what they paid.
For August 2023, Riot curtailed 84 GWh and collected $31 million for it. That works out to roughly 38 cents per kilowatt-hour — more than twice what Texas households pay, and about 14 times what Riot originally paid for the same electricity.
This is not a conspiracy. It is a contract, written and signed, operating exactly as intended for the benefit of the party that negotiated it.
Who Wrote the Rules
The Public Utility Commission (PUC) of Texas — the body that oversees ERCOT and sets the rules for demand-response payments — is appointed directly by the Governor of Texas, Greg Abbott. Abbott has publicly stated he wants Texas to be a Bitcoin "Mecca" and in June 2025 signed Senate Bill 21 into law, establishing the Texas Strategic Bitcoin Reserve. Public money is now held in Bitcoin as a state financial asset.
In August 2025, the Texas Attorney General ordered the release of data on cryptocurrency miners' power consumption and locations. The PUC sued the Attorney General to prevent that release. Its lawyers argued that disclosing miner energy data could enable terrorist attacks.
Data on Texas power plants — actual critical infrastructure — is publicly available. The "terrorism" argument has been reserved exclusively for Bitcoin mining facilities. The PUC is also refusing to release historical data from prior years. In the Texas Tribune's coverage, an independent energy expert described Bitcoin mining as "the very definition of a parasitic load."
Auditing Riot Platforms: August 2025
You do not need access to confidential ERCOT data to estimate a miner's energy consumption. Three figures from Riot's monthly "Production and Operations Updates" are sufficient. Here is the calculation for August 2025.
- Average operating hash rate: The effective hash rate during the month. Reported in EH/s (1 EH = 1 million TH).
- Fleet efficiency: Measured in J/TH. Tells you how much energy each unit of hashing consumes.
- Deployed hash rate: The maximum hash rate available. Used to estimate curtailment.
Step 1: Convert hash rate to TH/s
August 2025 average operating hash rate: 31.4 EH/s = 31.4 million TH/s.
Step 2: Calculate average power draw
Multiply the hash rate by fleet efficiency. Note that 1 Watt = 1 Joule per second:
31.4 million TH/s × 21.0 J/TH = 659.4 million J/s = 659.4 MW.
Step 3: Calculate monthly energy consumption
- Per hour: 659.4 MWh
- Per day (24 h): ~15.8 GWh
- For August (31 days): ~490 GWh
How much was curtailed
Repeating the calculation with the deployed hash rate gives the maximum possible consumption — what Riot could have used had nothing been curtailed. For August 2025, that comes to 568.7 GWh.
The difference, 78.7 GWh, is the upper bound on curtailment. The actual figure may be lower because some downtime also comes from facility maintenance, which Riot does not separately report.
What that energy was worth
Riot earned $16.1 million by curtailing at most 78.7 GWh. That is a sell-back price of at least $0.20/kWh. Their "All-in Power Cost — Total" for the month was $0.026/kWh. They sold the same energy back for at least 7.7 times their purchase price. If actual curtailment was lower than 78.7 GWh, the multiple was higher — at 65 GWh actual curtailment, the sell-back rate works out to roughly 9.5 times the purchase price.
The link between sell-back price and mining profitability
The August 2025 sell-back rate ($0.20/kWh) was substantially below the August 2023 rate ($0.38/kWh). The reason is mining profitability: hashprice was around $55 per PH/s in August 2025, compared to roughly $80 per PH/s in August 2023. When mining pays more, miners demand more to stop. The sell-back price is decided only by the opportunity cost of not mining bitcoin, it has nothing to do with the affordability of electricity for residents.
Note: Riot stopped publishing monthly figures in 2026, citing the same "terrorism risks" argument used in the PUC lawsuit. Quarterly estimates are still possible.
The Human Cost
In 2024, more than 3 million Texas households had their electricity disconnected for non-payment — the highest of any U.S. state, according to a May 2026 federal report.
Texas also hosts the world's highest concentration of Bitcoin mining — at least 27 facilities. It is the same state whose grid operators are publicly praised by the industry for "partnership" with miners. And it is the state where over 300 people died during the 2023 heatwaves because they could not afford to cool their homes.
All three facts are simultaneously true. The AI agent confirmed each of them individually. It could not explain how they fit together.
Where the Agent Stopped Engaging
Across the seven rounds I conducted, the agent acknowledged every alternative I proposed — dispatchable power plants, workload-shifting data centers (Google Cloud already does this for non-urgent compute), grid-scale storage, and China's grid, which is one of the world's largest and operates without any Bitcoin mining since the 2021 ban. The agent acknowledged each as legitimate. It did not explain why mining is preferable to any of them.
On the renewable energy claim, the agent defended that some miners run exclusively on surplus renewable energy and cited two examples by name: Gridless in Africa, and Crusoe Energy's flare-gas mitigation sites. The Crusoe citation is particularly worth noting, because Crusoe itself no longer agrees with the agent. In March 2025, Crusoe sold its entire Bitcoin mining operation — all 425 modular data centers, 270 MW of capacity, and 135 employees — to NYDIG. The reason, as Crusoe's co-founder Cully Cavness stated directly: "The AI business — it's become the majority of our revenue." The company the AI cites as proof that Bitcoin mining is the best use of stranded energy concluded that AI workloads were a better use of the same energy and exited mining entirely. This is the alternative I asked about in Round 2 of the dialogue — and it is now playing out at the scale of an entire company.
When I pointed out that hydropower is dispatchable (its "excess" only occurs during the rainy season), that solar excess only exists during daylight, and that wind is intermittent — meaning year-round profitable mining cannot run on excess renewables alone — the agent agreed and pivoted to arguing that Bitcoin mining makes renewable projects "financially viable" by buying their off-peak output. The fuller picture of how miners actually source power is in my earlier post on the renewable energy myth.
And the question I returned to — the one the agent never answered across all seven rounds — was:
If even miners can't afford electricity during price spikes, how will ordinary people — those living in orphanages and old age homes supported by donations — afford it during these spikes?
The agent answered with the Norway and Kenya cases. It claimed Norway's electricity prices rose after a Bitcoin miner left. The price rise was restricted to a small town in Northern Norway and was temporary; prices across Northern Norway are currently the lowest since 2020. It claimed Kenya's prices fell after miners joined the grid; its source was a single tweet, and Kenya's electricity prices are currently among the highest in Africa.
The affordability question was never directly addressed. The agent kept returning to the same handful of names and figures — Brad Jones, the $18 billion peaker plant figure, the 95% curtailment capability, July 2022 — regardless of what was being asked. When the same five examples answer every objection, even after each is rebutted, that is a signal about the depth of the supporting evidence, not its strength.
What This Means
Grid stabilization by Bitcoin miners is a real phenomenon. It is also conditional on a profit calculation that points the right direction. The contract structure — buying power at industrial rates and selling it back at peak rates — works as a commercial arrangement. It does not work as a public service, because public services do not stop functioning when private profit margins compress.
The deeper issue is one I have written about in my analysis of Bitcoin's broader extraction model and in why Proof-of-Work consumes the energy it does: the system is not designed to produce anything beyond its own continuation. When that continuation depends on competing with elderly residents for kilowatt-hours during a heatwave, the design priorities become visible.
For more on why Bitcoin has not delivered on its monetary promises despite this energy consumption, see the six essential jobs of money and why Bitcoin fails at each, and on the limits of its scaling solutions, my critique of the Lightning Network.
About the author: Pratik V. Padghane is an electrical engineer and science writer based in India. He is the author of Confessions of Bitcoin's AI: On Energy, the Grid & the Human Cost, which documents the full seven-round dialogue with annotated analysis, contractual breakdowns, and a comparison with Finland's community-first energy contract model. Read about me here.
No hype. No speculation. Just data-driven analysis.
Comments
Post a Comment