Three Midsize Companies Riding the AI Wave Into 2026

Fast Company has named three midsize companies — each employing between 100 and 499 people — among its Next Big Things in Tech honorees for 2026, spotlighting businesses that either streamline everyday work with AI or build the infrastructure AI itself demands.

The three winners span corporate event planning, chip-to-chip networking, and grid-scale energy storage. What they share is a bet that AI cuts both ways: it automates tedious tasks, and it creates urgent new needs for compute power and reliable electricity.

What does BoomPop bring to a $1.5 trillion market?

Corporate events are a $1.5 trillion market, yet planning them has historically left companies juggling spreadsheets, emails, and vendor websites. BoomPop attacks that friction with an AI-native platform that finds locations for events from corporate retreats to customer dinners, gathers rates and quotes, and enables booking with a few clicks.

The platform draws on a proprietary database of roughly 1.5 million vendors. It can also craft invitations, track RSVPs, and integrate with calendar software. BoomPop has booked more than $60 million in business for venues over the past two years, working with clients such as Google, Netflix, Amazon, and Dick's Sporting Goods.

The revenue figure matters as much as the client logos: it shows AI-native workflow tools can convert vendor-side bookings at scale in an industry still dominated by manual coordination.

How does Lightmatter speed up AI training?

Modern AI models rely on small neural networks called "experts" that focus on different types of problems and are typically spread across hundreds of GPU chips in a data center. Those chips need to share data, but electrical connections — and even traditional optical networks — cannot shuffle information fast enough.

Lightmatter's answer is Passage, a networking system powered by microscopic lasers. The company says the system lets models train up to three times faster and deliver results up to 11 times faster while using a fraction of previous power requirements.

In March 2026, Lightmatter and Qualcomm announced a record-breaking 1.6 Tbps bandwidth per fiber. Lightmatter has also launched an open standards initiative within the Open Compute Project aimed at setting standards for integrating fiber connections with computing chips — a move that could shape how the whole industry wires its data centers.

Why does Torus matter for the grid?

AI's energy appetite is colliding with an electrical grid built for steadier demand. Torus builds electrical storage systems that boost grid capacity and absorb sudden shifts in load.

Its Torus Station technology pairs the company's Torus Spin fast-responding mechanical flywheel systems with battery storage, responding to the grid's needs in milliseconds without degrading over time the way pure battery systems do. An AI orchestration platform called Torus Overwatch anticipates peaks and handles preventive maintenance, while built-in cybersecurity technology shields the system from hackers.

The company's 2025 results give the pitch substance:

• Helped utilities respond to more than 360 real-world demand response events
• Opened a 545,000-square-foot manufacturing facility in Salt Lake City
• Announced $200 million in funding from asset manager Magnetar

That combination of deployed hardware, new manufacturing capacity, and fresh capital positions Torus as demand for grid flexibility accelerates alongside AI data-center growth.

What ties the three together?

The honorees illustrate two sides of the same AI economy. BoomPop sells AI as a convenience, replacing tabbing between apps with a single platform. Lightmatter and Torus sell what AI consumes: faster interconnects and stable power. Fast Company selected the winners through its Next Big Things in Tech methodology, covering companies of all sizes across multiple categories.

For midsize companies watching these three, the lesson is concrete: the fastest-growing niches sit either in removing AI-era friction from routine business tasks or in supplying the energy and networking layers that AI's expansion now requires.