AI PCs & local-AI supercomputers: shipping now vs coming (2026)

A year ago, running a 70-billion-parameter model at home meant a tower of GPUs and a small space heater's worth of power draw. In 2026 it means a box the size of a paperback stack. The reason is unified memory: a single large pool the CPU and GPU both address, so a ~65–200 W desktop can hold a large model entirely in memory. That shift is what created the “AI PC” — and the pace of new hardware is fast enough to be worth tracking. Here's what's confirmed and shipping, and what's still a rumor.

Shipping now

• NVIDIA GB10 “DGX Spark”-class boxes are shipping. NVIDIA's DGX Spark went on sale in late 2025, and partner machines are now live from ASUS (the Ascent GX10), MSI (EdgeXpert), Dell, HP, Lenovo, Acer and Gigabyte. All pair a GB10 Grace-Blackwell chip with 128 GB of unified memory and NVIDIA's DGX OS / CUDA stack. Supply is constrained and pricing moves; the ASUS Ascent GX10 is the most affordable GB10 machine on Amazon.
• AMD Ryzen AI Max+ 395 “Strix Halo” is spreading fast. The 128 GB-capable APU now ships in the Framework Desktop, Beelink GTR9 Pro, Minisforum MS-S1 Max, GMKtec EVO-X2, HP Z2 Mini G1a, Corsair's AI Workstation and more — the value story for big local models. AMD has even published guidance on clustering multiple Ryzen AI Max nodes for trillion-parameter mixture-of-experts inference.
• Apple M5 shipped — in laptops. The M5 MacBook Air and M5 Pro/Max MacBook Pro arrived in early 2026. M5 desktops (Mac mini, Mac Studio) are not out yet, so the current desktop local-LLM Macs remain the M4-family Mac mini and the M4 Max / M3 Ultra Mac Studio.
• Intel Panther Lake (Core Ultra Series 3) launched at CES 2026 — Intel's first 18A-process mobile chip, with an upgraded NPU — pushing the thin-and-light “AI PC” laptop tier forward (though, as ever, the NPU matters far less than memory bandwidth for large local models).

On the horizon (rumor / roadmap)

• NVIDIA RTX Spark (rumored, Computex 2026). A consumer Windows-on-Arm PC platform co-designed with Microsoft — reportedly up to 20 cores, an RTX 5070-class GPU and up to 128 GB — with laptops said to arrive around Fall 2026 from Dell, ASUS, HP, Lenovo, Microsoft and MSI, starting near $3,000. It would be NVIDIA's push into mainstream Windows PCs. Specs and pricing are unconfirmed.
• AMD “Medusa Halo” (leak, ~2027). The rumored Strix Halo successor — Zen 6 plus RDNA 5, with a leaked 384-bit LPDDR6 bus said to raise memory bandwidth by roughly 80%. More bandwidth is exactly what these machines need most. Leak-level only; treat the numbers as unconfirmed.
• NVIDIA “Vera Rubin Spark” (roadmap, 2027–2028). NVIDIA's own published roadmap points to a GB10 successor on LPDDR6, with lower-cost variants planned — but that's availability years out, not a 2026 purchase.
• Apple M5 Ultra / M5 Mac desktops (expected, unannounced). An M5 Mac mini and Mac Studio are widely expected in late 2026, possibly slipping into 2027. Unannounced as of July 2026.

At a glance: now vs next

Don't want to wait? What to buy today

If you want to run local models now rather than wait on a rumor, one machine per silicon family covers the range — all verified on Amazon at research time (re-check the seller and price, which move on these niche systems):

• ASUS Ascent GX10 — NVIDIA GB10 · CUDA stack — ~$3,971. Check price on Amazon →
• Beelink GTR9 Pro — Ryzen AI Max+ 395 · 128 GB value — ~$1,999. Check price on Amazon →
• Apple Mac mini (M4 Pro) — Apple M4 Pro · turnkey — ~$1,599. Check price on Amazon →

Full picks, verdicts, pros/cons and a speed/memory comparison are in our best AI PC for local LLMs guide and the AI PC benchmark.

Frequently asked questions

Sources

Confirmed specifications and shipping status are drawn from manufacturer pages: NVIDIA's DGX Spark (GB10, 128 GB unified, 273 GB/s) and Apple's Mac Studio and Mac mini spec pages for Apple memory bandwidth. AMD Ryzen AI Max (“Strix Halo”) specifications and the mini-PC landscape are reported by ServeTheHome, Tom’s Hardware and NotebookCheck; independent local-LLM throughput figures come from ServeTheHome, Level1Techs and LMSYS. The rumored items (RTX Spark, AMD Medusa Halo, Vera Rubin Spark, M5 desktops) are leak- or roadmap-level, attributed to the hardware press and labelled as unconfirmed above. This page was produced with AI assistance; every confirmed spec was cross-checked against a primary source before publication.

Related buyer's guides

All reviews →

Buyer's guide / PC builds

Best gaming PC under $1,500 in 2026

The honest $1,500 build: a 16GB RX 9060 XT, Ryzen 5 and 32GB DDR5 for strong 1080p-ultra and entry-1440p gaming — with the real shortage-adjusted total, where every dollar goes, and whether a prebuilt beats it this year.

See the build →

Buyer's guide / PC builds

Best gaming PC under $2,000 in 2026

The value sweet spot: near-flagship 1440p gaming built around the RX 9070 XT and an X3D CPU — with the honest 2026 total, budget allocation, and exactly how to duck under a hard $2,000.

See the build →

Buyer's guide / PC builds

Best gaming PC under $3,000 in 2026

Where 4K becomes real: the fastest gaming CPU paired with an RTX 5080 for 1440p-max and 4K-high with DLSS 4 — the honest shortage-adjusted total, where every dollar goes, and a matched monitor.

See the build →

Buyer's guide / PC builds

Best gaming PC under $5,000 in 2026

The honest high-end build: a fully-maxed RTX 5080 creator-gaming machine with a 16-core 9950X3D, 64GB and 4TB — plus the truth about why $5,000 does not reliably buy a 5090 at 2026 street prices.

See the build →

Buyer's guide / AI PCs

LLM VRAM calculator

Enter a model size, quantization and context length to estimate the VRAM or unified memory an LLM needs to run locally — and instantly see which 2026 machines can hold it.

Use the calculator →

Buyer's guide / AI PCs

How to run LLMs locally: the 2026 hardware guide

How much VRAM or unified memory you need to run 7B-120B models locally, the capacity-vs-bandwidth tradeoff, and the 2026 hardware that hits each point on the curve — GB10, Ryzen AI Max, Apple Silicon and RTX GPUs.

Read the guide →