Inertia Enterprises has raised $450 million to build one of the world’s most powerful lasers, which it hopes will serve as the foundation of a grid-scale power plant the fusion startup intends to start construction on in 2030. Inertia Enterprises is building on technology developed at the Lawrence Livermore National Laboratory’s National Ignition Facility (NIF). The NIF is the site of the world’s only controlled fusion reactions that have reached scientific breakeven, in which the reaction releases more energy than it took to start. 

The Series A was led by Bessemer Venture Partners with participation from GV, Modern Capital, Threshold Ventures, and others. Inertia’s co-founders include Jeff Lawson, who co-founded Twilio and serves as its CEO; Annie Kritcher, who led the successful experiments at NIF; and Mike Dunne, a Stanford professor who helped Lawrence Livermore develop a power plant design based on NIF. Kritcher has remained in her position at Lawrence Livermore.

NIF’s breakeven experiments have been a key milestone on the road to widespread fusion power. However, considerable progress needs to be made before a fusion power plant can deliver electricity to the grid. For Inertia, that means building a laser capable of delivering 10 kilojoules 10 times per second.

The startup’s reactor relies on a form of fusion known as inertial confinement. In Inertia’s flavor of inertial confinement, lasers bombard a fuel target, compressing the fuel until atoms inside fuse and release energy. The technique is based on NIF’s designs, in which laser light is converted into X-rays inside the target. The X-rays are what ultimately heat and compress the fuel pellet.

Each of Inertia’s power plants will require 1,000 of its lasers bombarding 4.5 mm targets that cost less than $1 each to mass produce. By contrast, the NIF’s system uses 192 lasers to fire on painstakingly crafted targets that take dozens of hours to make. Inertia is betting that by using the same basic principles as NIF and applying a more commercial mindset, it can bring the costs down dramatically.

Inertia’s new round is the latest in a string of funding announcements from fusion startups in recent months. With this round and others, fusion startups have attracted more than $10 billion in investments. And at least a dozen companies have raised more than $100 million.

Last week, Avalanche said it had raised $29 million to advance its desktop-sized fusion reactor. Earlier this year, Type One Energy told TechCrunch it had attracted $87 million in investment in advance of a $250 million Series B that it’s currently raising. Last summer, Commonwealth Fusion Systems raised $863 million from dozens of investors, including Google, Nvidia, and Breakthrough Energy Ventures.

Two fusion companies recently announced they were going public via reverse mergers. General Fusion said in January it would merge with acquisition company Spring Valley III in a deal that values the combined company at $1 billion. General Fusion had previously struggled to raise money from private investors. Earlier last month, TAE Technologies announced it would merge with Donald Trump’s social media company, Trump Media & Technology Group; the combined company would be worth $6 billion, according to the all-stock transaction.

UpScrolled, a social network that caught fire after TikTok’s ownership change in the U.S., is facing a serious moderation problem. After growing to more than 2.5 million users in January, users have reported the app is not taking action on the creation of usernames and hashtags that contain racial slurs, and hasn’t been able to properly moderate harmful content.

After receiving tips from UpScrolled users, TechCrunch confirmed the existence of a wide range of racial slurs and hate speech being used in people’s usernames on the app. For instance, some usernames would feature the name of the slur itself, the slur combined with other words, or multiple slurs in a single username; other usernames contain hate speech, like “Glory to Hitler.”

After reporting these slurs to UpScrolled’s public email address, we received a response from that email that the company is “actively reviewing and removing inappropriate content,” and is working to expand its moderation capacity. The email advised us not to engage with bad-faith actors while the situation is resolved.

Days after reporting this activity on the app, the accounts with slurs in the usernames that were provided to UpScrolled via screenshots remained online.

In addition, slurs and hate speech can be found elsewhere in the app, including hashtags and text used alongside its photo or video content, TechCrunch found. Other harmful content was available, including text posts with racial slurs and hate speech, and photo and video content glorifying Hitler, based on TechCrunch’s review of the app.

TechCrunch wasn’t alone in identifying this problem; the ADL also published a blog post this month, noting that UpScrolled was becoming home to antisemitic and extremist content and designated foreign terrorist organizations, like Hamas and others.

UpScrolled, which was founded in 2025, claims on its website claims that the platform offers every voice “equal power.” The app has seen more than 4 million downloads on iOS and Android since June 2025, according to market intelligence provider Appfigures — a figure even higher than the startup’s self-reported number last month.

But while UpScrolled’s FAQ explains the app doesn’t “censor opinions,” it does indicate that its policy is to restrict content that involves “illegal activity, hate speech, bullying, harassment, explicit nudity, unlicensed copyrighted material, or anything intended to cause harm.”

That guidance is similar to most modern-day social media platforms. It’s clear, however, that the company is struggling to enforce its rules.

It’s battle that social networks are often faced with — especially those that receive a large influx of new users in a short time period. Bluesky, for instance, faced issues with slurs in account usernames in July 2023, which led to users threatening to leave the site.

After UpScrolled’s initial reply to our inquiry, TechCrunch also received a response from the press account on Tuesday, which directed us to UpScrolled founder Issam Hijazi’s new video, where he addressed the issues with content moderation.

For 24 years, Microsoft’s Amanda Silver has been working to help developers — and in the last few years, that’s meant building tools for AI. After a long stretch on GitHub Copilot, Silver is now a corporate vice president at Microsoft’s CoreAI division, where she works on tools for deploying apps and agentic systems within enterprises.

Her work is focused on the Foundry system inside Azure, which is designed as a unified AI portal for enterprises, giving her a close view of how companies are actually using these systems and where deployments end up falling short.

I spoke with Silver about the current capabilities of enterprise agents, and why she believes this is the biggest opportunity for startups since the public cloud.

This interview was edited for length and clarity.

So, your work focuses on Microsoft products for outside developers — often startups that aren’t otherwise focused on AI. How do you see AI impacting those companies?

I see this as being a watershed moment for startups as profound as the move to the public cloud. If you think about it, the cloud had a huge impact for startups because it meant that they no longer needed to have the real estate space to host their racks, and they didn’t need to spend as much money on the capital infusion of getting the hardware to be hosted in their labs and things like that. Everything became cheaper. Now agentic AI is going to kind of continue to reduce the overall cost of software operations again, because many of the jobs involved in standing up a new venture — whether it’s support people, legal investigations — a lot of it can be done faster and cheaper with AI agents. I think that’s going to lead to more ventures and more startups launching. And then we’re going to see higher-valuation startups with fewer people at the helm. And I think that that’s an exciting world. 

What does that look like in practice?

We are certainly seeing multistep agents becoming very broadly used across all different kinds of coding tasks, right? Just as an example, one thing developers have to do to maintain a codebase is stay current with the latest versions of the libraries that it has a dependency on. You might have a dependency on an older version of the dot-net runtime or the Java SDK. And we can have these agentic systems reason over your entire codebase and bring it up to date much more easily, with maybe a 70% or 80% reduction of the time it takes. And it really has to be a deployed multistep agent to do that.

Live-site operations is another one — if you think of maintaining a website or a service and something goes wrong, there’s a thud in the night, and somebody has to be on call to get woken up to go respond to the incident. We still do have people on call 24/7, just in case the service goes down. But it used to be a really loathed job because you’d get woken up fairly often for these minor incidents. And we’ve now built a genetic system to successfully diagnose and in many cases fully mitigate issues that come up in these live site operations so that humans don’t have to be woken up in the middle of the night and groggily go to their terminals and try to diagnose what’s going on. And that also helps us dramatically reduce the average time it takes for an incident to be resolved.

One of the other puzzles of this present moment is that agentic deployments haven’t happened quite as fast as we expected even six months ago. I’m curious why you think that is.

If you think about the people who are building agents, what is preventing them from being successful, in many cases, it comes down to not really knowing what the purpose of the agent should be. There’s a culture change that has to happen in how people build these systems. What is the business use case that they are trying to solve for? What are they trying to achieve? You need to be very clear-eyed about what the definition of success is for this agent. And you need to think, what is the data that I’m giving to the agent so that it can reason over how to go accomplish this particular task?

We see those things as the bigger stumbling blocks, more than the general uncertainty of letting agents get deployed. Anybody who goes and looks at these systems sees the return on investment.

You mention the general uncertainty, which I think feels like a big blocker from the outside. Why do you see it as less of a problem in practice?

First of all, I think that it’s going to be very common that agentic systems have human-in-the-loop scenarios. Think about something like a package return. It used to be that you would have a workflow for the return processing that was 90% automated and 10% human intervention, where somebody would have to go look at the package and have to make a judgment call as to how damaged the package was before they would decide to accept the return. 

That’s a perfect example where actually now the computer vision models are getting so good that in many cases, we don’t need to have as much human oversight over inspecting the package and making that determination. There will still be some cases that are borderline, where maybe the computer vision is not yet good enough to make a call, and maybe there’s an escalation. It’s kind of like, how often do you need to call in the manager? 

There are some things that will always need some kind of human oversight, because they’re such critical operations. Think about incurring a contractual legal obligation, or deploying code into a production codebase that could potentially affect the reliability of your systems. But even then, there’s the question of how far we could get in automating the rest of the process.