In a recent interview with Lesley Stahl on 60 Minutes, Donald Trump changed his tone on climate change—sort of.

“I’m not denying climate change,” he said. “But it could very well go back.”

Scientists overwhelmingly agree that President Trump is wrong: To suppose that climate change is somehow reversible in the natural course of events on any meaningful timescale is tantamount to denial. The science states clearly that the choices made in climate policy in the coming decades will likely affect the planet for millennia. Nothing will be reversible, except perhaps on geologic time scales—tens of thousands of years, by which time whatever remains of Trump’s towers will likely be under literal, if not financial, water.

Still, there are limits to our understanding about climate change in the very long term. That’s because climate models tend to aim at the year 2100, because of both practical and technological limitations on what can be modeled. After that, we’re looking at very long-term timescales. But no matter how we model, the effects will clearly be irrevocably catastrophic unless emissions are reduced, and fast.

On October 8, the United Nations’ Intergovernmental Panel on Climate Change (IPCC), the gold standard of climate science, released a special report, focusing on the effects of 1.5 degrees Celsius of warming. It’s a particularly dire one, even given the subject: The 2 degrees Celsius of global warming above preindustrial levels, once considered the limit for avoiding catastrophe, is now recognized as having catastrophic consequences. Wildfires, coral reef die-offs, droughts and floods will all get worse by 2040. And ambitious geoengineering tech that doesn’t exist yet could be crucial in mitigating some of warming’s worst effects.

But even these reports have to stop somewhere, and the date is 2100, when most of the IPCC’s climate projections end. The reason for the cut-off is simple: computing power and time.

When the IPCC started its work in the late 1980s and early ’90s, the computers simply didn’t have the capacity to run models that reached beyond 2100 in a reasonable amount of time.

“It was a technological limitation, but it kind of stuck,” said Peter Clark, a climatologist at Oregon State University, and an author of the 2013 IPCC report’s chapter on sea-level rise.

The models that have been developed since have grown increasingly sophisticated, but still largely take aim at the end of the century. That explains why the reports from the IPCC have so far only hinted at what might happen after 2100. Running these models on tennis-court-sized supercomputers uses up a ton of energy, too. A dismal side effect of figuring out how bad things will be in a future climate is that it takes a lot of power to do it.

The effects on specific regions through 2100, how the melting of certain ice sheets will affect regions of the world, the detail of the different emissions scenarios—all that has gotten much more granular that it used to be through the year 2100. But not after.

Of course, it doesn’t take a supercomputer—or even a clairvoyant—to imagine what happens next. Neither the cataclysmic events of climate change nor warming will stop in 2100, even if emissions do miraculously taper off in time to level off warming at 1.5 or 2 degrees Celsius above preindustrial levels. (Those are the levels deemed admissible by the Paris Agreement.)

There’s a lag between emissions and warming: any carbon that’s currently being pumped into the atmosphere will take decades to warm up the world. Research suggests that up to nearly half of the world’s ice sheets can’t handle our current climate; they’d still melt if emissions stopped today. Seas will continue to rise long after 2100, even under a 1.5 degree warming scenario.

“What we’re doing right now is we’re putting a big spike of carbon into the atmosphere—instantaneously on geologic time scales,” said Anji Seth, a climate scientist at the University of Connecticut. “Even 50,000 years in the future there will still be more carbon in the atmosphere than there was before.”

Still, few models have so far aimed to figure out just how much—but that might be a good thing, because it pushes people to think more practically.

“The number-one reason why we’re doing future simulations is to model the choices that we still can make today,” said Katharine Hayhoe, a climate scientist at Texas Tech University. The further out the projections, the less connected they are to the emissions we can still control.

The IPCC nevertheless is “virtually certain” that sea-level rise will continue beyond the end of the century. October’s special report mentions that on very long timescales, up to 50 meters of sea-level rise is possible. As 2100 inches closer, the scientific community is training its attention increasingly on the longer-term effects of climate change.

Paleoclimatologists, who study climates in our distant past, are working with scientists who model future scenarios to figure out how the world’s climate can change over long timescales. The next big IPCC report, due in two to three years, will dig into the more permanent effects of our current emissions.

There are still technological limitations to modeling farther into the future, researchers said. “Every single terabyte on the servers models to the year 2100,” said Benjamin Horton, a paleoclimatologist at the Earth Observatory of Singapore, and an editor on the IPCC’s upcoming sixth assessment. “I know this only too well because I ask them if they will go back in time for me, and they sometimes won’t.”

Computing power is still ultimately finite, too, so scientists still have to choose which models to run, while we still have time not to completely char the planet.

“Now we have to make choices between higher resolution (which you need if you are going to provide fine scale information, for which there is a great need), more runs (because you want to simulate the fact that the climate system has intrinsic variability), and using a lower resolution model for extremely long time scales (a thousand, or even thousands of years),” Susan Solomon, a professor of environmental science at MIT said in an e-mail. “All of these are necessary to get different kinds of information.”

Those studies that have reached beyond the 21st century indicate that the seas will rise for millennia because of our emissions in just the next few decades. If temperatures top out at 3 degrees Celsius above preindustrial levels, research suggests that between two and three dozen countries, as they currently exist, could lose over 10 percent of their land over the course of the next two millennia.

It’s easier to heat the planet than it is to cool it, though. “It’s relatively easy once you’ve started the process to collapse parts of the ice sheets,” Horton said. “It’s very difficult to grow an ice sheet. Our climate tends to warm up fast and cool down slow.”

What that means, practically, is that with the warming that’s already been put into motion, we’re very likely looking at a sustained increase in sea levels for tens of thousands of years. The only question is how fast the waters will rise. And that, once again, depends entirely on how much carbon we emit in the coming decades while we can still put mitigation efforts into place, by building seawalls, moving away from the coasts, and protecting fragile coastal ecosystems.

“Let’s just look out the full 10,000 years,” Clark told me over the phone. “In the low-end scenario that we considered, you’re looking at 25 meters of sea-level rise. In the high-end scenario, you’re looking 55 meters of sea-level rise.” That rise will happen pretty quickly, considering the timescale, mostly in the next 2,000 to 3,000 years, according to Clark’s research.

After all, humans will use up the vast majority of our fossil-fuel reserves in the next century if we continue on the current trajectory. We can only emit so much carbon.

That means there’s not much time to avoid a catastrophic—and, unless we’re talking in geologic time—permanent fate for our coastlines. “The trajectory that we end up on over the course of the next several millennia is going to be decided largely in the next couple of decades,” Clark said.

In his conversation with Lesley Stahl, Trump went on to suggest that climate science often refers to distant projections—an idea he invoked, purportedly, to justify his own denialism.

“You know, we’re talkin’ about over a millions of years,” Trump told Stahl. Indeed we are.