The Guardian's reporting frames the crossover as a structural milestone driven by EIA data showing coal at ~14% of US electricity generation while solar surpassed it. The article emphasizes that coal capacity has shed roughly 100 GW since 2015 because plants retired faster than utilities planned — the economics simply stopped working regardless of political headwinds.
Argues coal didn't lose to policy but to a levelized cost curve that kept bending while coal's stayed flat. Points to 240 GW of utility-scale solar plus another ~50 GW of behind-the-meter rooftop as evidence that market forces, not regulation, drove the crossover.
Highlights that 12 GW of new battery capacity was deployed in the past 12 months, mostly paired with solar in ERCOT and CAISO. Argues this transforms solar from a daytime-only resource into something dispatchable across the evening peak — precisely the demand window coal historically defended.
Frames the milestone as directly relevant to engineering decisions because US electricity demand is growing for the first time in two decades, driven overwhelmingly by AI compute buildouts in Northern Virginia, Phoenix, Columbus, and Dallas. Argues every announced GW of AI training capacity has a power-sourcing story behind it that is increasingly solar-plus-storage, making carbon intensity per token a real procurement metric.
Submitted the Guardian article to Hacker News where it drew 443 points and 209 comments — signaling that the developer community treats the solar-coal crossover as technically and economically relevant, not merely an environmental headline.
For the first time in US history, solar generated more electricity than coal over a sustained reporting window. The Guardian's June 11 report, citing EIA data, marks the crossover as a structural one: solar's lead isn't a seasonal spike or a single-month artifact, it's a trend line that finally crossed coal's declining one. Coal generated roughly 14% of US electricity in the period covered; solar cleared that bar.
The numbers behind the headline are worth pausing on. Utility-scale solar capacity in the US passed 240 GW earlier this year, with another ~50 GW of behind-the-meter rooftop. Coal capacity, by contrast, has shed about 100 GW since 2015 — plants retired faster than utilities planned, because the economics stopped working. Coal didn't lose to policy. It lost to a levelized cost curve that kept bending while coal's stayed flat.
The other quiet story is storage. Battery deployments hit a record 12 GW of new capacity in the past 12 months, mostly paired with solar in ERCOT and CAISO. That changes solar from a daytime-only resource into something dispatchable across the evening peak — which is precisely the window coal used to defend.
For a developer audience, the temptation is to file this under "climate news, not my problem." That's wrong, and the reason is datacenters.
US electricity demand is growing again — for the first time in two decades — and the marginal new load is overwhelmingly compute. Hyperscaler buildouts in Northern Virginia, Phoenix, Columbus, and the Dallas crescent have pushed grid operators into capacity warnings they hadn't issued since the 1990s. Every GW of AI training capacity you've heard announced in the last 18 months has a power-sourcing story behind it, and that story is increasingly solar-plus-storage, not coal.
This matters for three reasons that show up directly in engineering decisions.
First, carbon intensity per token is now a real procurement metric. Anthropic, Google, and Microsoft all publish compute-region carbon data, and enterprise buyers have started filtering on it. A region running on a 70% solar mix at midday has a fundamentally different gCO2/kWh number than one drawing from a coal-heavy ISO. If you've been ignoring `AWS_REGION` as a sustainability lever, your procurement team probably hasn't.
Second, interconnect queues have replaced capex as the bottleneck. The reason solar overtook coal isn't that anyone built coal — it's that solar got built. Roughly 2,600 GW of generation and storage projects are sitting in US interconnect queues right now, with solar and batteries the dominant share. The constraint on new datacenter siting in 2026 isn't "can we get a permit" — it's "can we get a substation slot before 2029." That's why hyperscalers are signing 15-year solar PPAs in regions they don't even operate in yet: they're locking in queue position.
Third, the economic case is now unsentimental. Lazard's most recent LCOE work puts unsubsidized utility solar in the $29–38/MWh range and new coal at $70–168/MWh. Those numbers are not close. Even with the storage adder, solar-plus-4-hour-battery comes in under $80/MWh in most of the sunbelt. Coal isn't being phased out; it's being outbid in every capacity auction it shows up to.
The community reaction on Hacker News (443 points at writing) split predictably between people pointing out that solar's capacity factor is ~25% vs coal's ~50% — true — and people pointing out that nobody is building coal anyway, which is also true. The capacity-factor gap is real but solved by overbuild plus storage, which is what's actually happening.
If you operate infrastructure, this changes a few things you can act on this quarter.
Region selection is now a multi-variable optimization. Latency and price are the obvious axes; grid carbon intensity is the third, and it's increasingly the one your customers ask about in security reviews. AWS's `us-west-2` and Google's `europe-west1` have had hours where they ran near-100% renewable in 2026. Your inference workload doesn't care which rack it lands on — your sustainability report does.
Time-shifted compute is no longer hypothetical. When grid mix swings from 60% solar at noon to 30% solar at 8pm, batch jobs that can wait have a measurable carbon delta. Tools like Google's Carbon-Intelligent Compute and the open-source `electricitymap-api` make this actionable in a few hundred lines of scheduler code. The cheapest hour to run a training job is increasingly the cleanest one, which is a coincidence that won't last forever — but it's true today.
Hyperscaler PPAs are now a procurement signal you can read. When AWS announces a 1.2 GW solar PPA in West Texas, what they're really announcing is where their next-generation training clusters will land. Following the PPA announcements is a better leading indicator of region capacity than the AWS regional roadmap blog.
The other practical change: on-prem and colo customers in coal-heavy ISOs (MISO, PJM's western footprint) are starting to see rate increases as coal plants retire and the replacement capacity is built into ratebases. If your colo contract is in Ohio, Indiana, or West Virginia, the next renewal is going to hurt. Plan accordingly.
The coal-to-solar crossover is the kind of milestone that looks inevitable in retrospect and looked impossible ten years ago. The deeper signal is what it implies about the next crossover — solar versus gas — which is now a 4-to-6 year question rather than a 20-year one. For practitioners, the actionable read is simpler: the power map under your workloads has been redrawn, and the regions that win the next decade of compute are the ones that wired up to the sun before everyone else figured out the math.
The growth of solar is astounding. I dug into data a while back and tried to do some visualizations of it, mainly for my own understanding:https://torkeldanielsson.se/solar-energy-forecasts/Solar is already by far our cheapest source of energy. As solar expands, the learning rate
+1 to the Guardian for mentioning their data source, but -1 for not linking to it.+2 for EMBER for having a data source AND being able to link to the parameters that show solar overtaking coal for the month in the US.https://ember-energy.org/data/electricity-data-explorer/?e
Question for those in the know... See lots of press about balcony solar in Germany, and California recently introduced a bill to allow it (I'm guessing other states already allow it; not sure if the CA bill has a chance of becoming law). But how far are we from a more plug and play home solar s
The US currently is at per capita GHG emissions approximately at the the same level as it was in 1910.https://ourworldindata.org/profile/co2/united-statesDespite not being in the paris treaty, the us needs only a 10-12% reduction to meet the paris accord requirements on sche
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
This is more from a lot of coal power plants being converted to gas over the past 20 years than solar overtaking the outputs of those power plants. Coal output shrinking, solar output rising, the lines have crossed.Coal is unpopular in all but a few areas where coal mining is still a part of the loc