The artificial intelligence (AI) boom has rewarded Nvidia (NVDA 3.92%) shareholders handsomely, but the infrastructure buildout creating those gains demands more than just graphics processing units (GPUs). Data centers require faster optical connections, cleaner power delivery, a reliable nuclear baseload, specialized software platforms, and the manufacturing equipment that enables the creation of advanced chips.
These eight stocks tackle different layers of the AI stack. Some are proven businesses with decades of operating history. Others are precommercial companies with technologies that may or may not reach scale. Here's why I'd buy all eight right now.
Image source: Getty Images.
The photonics breakthrough
Poet Technologies (POET 5.45%) builds optical engines based on its Optical Interposer silicon-photonics platform, targeting 800-gigabit and 1.6-terabit data center links. Partnerships with Foxconn, Semtech, and Sivers Semiconductors validate the technology, and a $150 million capital raise in October 2025 strengthens liquidity through at least 2026. The optical interconnect market could grow to between $26 billion to $35 billion by 2030, and Poet's integrated approach eliminates multiple components that traditional solutions require.
NASDAQ: POET
Key Data Points
The power semiconductor pivot
Navitas Semiconductor (NVTS 8.33%) supplies gallium nitride (GaN) and silicon carbide (SiC) power semiconductors that support Nvidia's 800 VDC AI data center architecture. The "Navitas 2.0" strategy prioritizes high-power markets led by AI data centers. GaN and SiC significantly reduce conversion losses, with full-system efficiency gains of approximately 5% and certain stages achieving even higher reductions. Even small efficiency gains can save large data center operators millions of dollars annually in electricity costs, creating recurring demand as AI infrastructure scales up.
The nuclear power play
Oklo (OKLO 6.13%) develops small modular reactors. The company has a 12-gigawatt (GW) long-term power purchase agreement with Switch Data Centers running through 2044, and additional memoranda of understanding push its potential pipeline above 14 GW. Oklo broke ground in September 2025 on its first commercial reactor at Idaho National Laboratory, targeting initial operations in late 2027 to early 2028. As AI data centers demand clean, reliable baseload power, Oklo's colocation model addresses both energy stability and grid strain.
The enterprise AI platform
Palantir Technologies (PLTR 5.27%) operates government and enterprise AI software platforms, including Foundry, Gotham, and the Artificial Intelligence Platform (AIP). In Q3 2025, Palantir reported roughly $1.2 billion in revenue, marking its first quarter at that level. The company has turned its interest in AI into recurring enterprise contracts, demonstrating that corporations are willing to pay premium prices for software that delivers measurable productivity gains, making its stock a must-own AI play right now.
NASDAQ: PLTR
Key Data Points
The chip equipment cornerstone
ASML Holding (ASML 1.85%) manufactures extreme ultraviolet (EUV) and deep ultraviolet (DUV) lithography systems, which are critical to producing leading-edge chips that power AI and high-performance computing. ASML effectively holds a global monopoly on EUV technology, supplying Taiwan Semiconductor Manufacturing Company, Intel, and Samsung. Each EUV system costs between $180 million and $380 million, depending on the configuration. The company's installed base of more than 1,600 lithography tools generates recurring service and upgrade revenue, providing long-term stability.
The data center builder
Applied Digital (APLD 10.11%) designs and operates purpose-built data centers optimized for AI and high-performance computing workloads. Its Ellendale, North Dakota, campus is planned to reach 400 megawatts (MW) of capacity when fully built out. In January 2025, Macquarie Asset Management committed up to $5 billion to support Applied's AI infrastructure expansion. A 400 MW hosting agreement with CoreWeave underscores strong demand. The company sources low-cost stranded or renewable energy -- often at rates of around $0.02 to $0.03 per kilowatt-hour -- and leases capacity to AI clients at significantly higher rates, thereby creating wide operating margins.
The Nvidia alternative
Advanced Micro Devices (AMD 1.36%) competes directly with Nvidia in AI accelerators through its MI300X and MI300A series GPUs, built on the CDNA 3 architecture, alongside its Epyc data center central processing units. Major customers, including Microsoft, Meta, Oracle, and Dell, have announced deployments or partnerships, validating AMD's expanding foothold. Hyperscalers are pursuing multivendor GPU sourcing to avoid dependence on Nvidia, and AMD is capturing that demand as Nvidia's supply remains constrained.
NASDAQ: AMD
Key Data Points
The service robotics upstart
Richtech Robotics (RR 5.92%) designs AI-powered service robots for restaurants, hotels, hospitals, and industrial facilities, with more than 400 units deployed globally. In mid-2025, the company announced a $4 million sales agreement in China and began integrating Nvidia's Jetson Thor AI platform to enhance autonomous navigation and perception. Ongoing labor shortages in hospitality and healthcare continue to drive adoption, while the company's Robot-as-a-Service model typically achieves customer payback within 12 months.
