Question 1

Why Apple Silicon, and why now?

Accepted Answer

Apple has quietly shipped the largest fleet of AI-ready hardware in the world. There were 2.5 billion active Apple devices in January 2026, and every iPhone, iPad and Mac on Apple Silicon pairs a GPU and a Neural Engine with one shared pool of fast unified memory.

While others pour hundreds of billions into datacenters, Apple has largely sat the race out. Call it the last mover. Its bet is different: run what you can on the device, and send only what you must to the cloud. MLX connects that fleet into one network, owned by the people who power it.

Apple Inc. · AAPL

—

Loading…

Log scale. Split-adjusted closing prices. Shown for context, not as a prediction or advice.

Sources: 9to5Mac, Jan 2026

Question 2

Why would this work where GPU marketplaces stalled?

Accepted Answer

A compute network is a marketplace, and every marketplace starts the same way: it needs supply before anyone buys, and buyers before supply pays. MLX starts on the supply side with hardware people already own: Macs that sit idle most of the day.

Joining is one download, every Mac that works is rewarded from its first hour, and the same people who lend their Macs are the ones who use AI. Nobody can say today what millions of linked Macs will be used for in 2028; the network is built to find out.

Question 3

Who uses the compute?

Accepted Answer

The same people who lend their Macs are the ones who use AI. Today Macs mine MLX for verified jobs, and the Compute API (a pilot) pays USDC to the Macs that ran a job; paid AI jobs between Macs are planned. Nobody can say today what millions of linked Macs will be used for in 2028; the network is built to find out. None of this is a promise of demand, revenue or a price for MLX.

Question 4

Why are Nvidia, OpenAI, Anthropic and Google racing to build datacenters?

Accepted Answer

Because so far, more compute has meant better models. Training and serving frontier AI takes enormous clusters of GPUs, and whoever controls the most compute sets the pace. The numbers are hard to picture. Amazon, Microsoft, Alphabet and Meta planned more than $700 billion of capital spending for 2026, most of it for AI datacenters. Nvidia, which sells most of the chips, is worth around $5.7 trillion. Money isn't the only limit. Every new datacenter needs gigawatts of power, cooling, land, chips and years of building. The devices people already own need none of that: they're bought, powered and connected.

Question 5

Apple Silicon or Nvidia: which wins?

Accepted Answer

Chip for chip, Nvidia, by a wide margin. A datacenter GPU like the H100 is built for dense AI math and wired to thousands of others with very fast links. The largest frontier models will keep training in places like that. Apple Silicon wins on different terms: memory and efficiency. A Mac Studio with M3 Ultra holds up to 512 GB of unified memory, enough to run a model with more than 600 billion parameters on one desk. An H100 has 80 GB. Apple chips also use a fraction of the power. So it isn't either-or. Datacenters train the giants. A network of Apple devices runs models, fine-tunes them and handles everything that should stay private.

Question 6

Can millions of devices really work as one supercomputer?

Accepted Answer

It has happened before. In 2020, Folding@home volunteers passed an exaflop of combined power on their home computers, more than the fastest supercomputers of the time. A network like this suits work that splits into independent pieces: answering prompts, batch jobs, embeddings, evaluations and fine-tuning. Research into low-communication training, such as DeepMind's DiLoCo, keeps narrowing the gap for training too. It doesn't suit jobs where every chip must talk to every other chip thousands of times a second. MLX is designed around that line, routing each task to where it runs best.

Question 7

What could the largest Apple Silicon network be worth?

Accepted Answer

Nobody knows, so here is a thought experiment. Instead of guessing a price, ask what it would cost to buy the same capacity as datacenter GPUs. Drag to size the network.

Question 8

Local or cloud: why does hybrid win?

Accepted Answer

Today you mostly choose between a closed model in someone else's datacenter, which sees everything you send it, and a small local model that keeps your data private but can do less. Hybrid removes that choice. Your personal context, like messages, files, photos and health data, stays on your device, where a local model reads it. Work that needs more muscle goes out as a narrow, stripped-down request to the network or to a frontier model. Apple built Apple Intelligence on the same idea, pairing on-device models with Private Cloud Compute. Picture one assistant that drafts on your iPhone, thinks on the largest Apple Silicon supercomputer in the world, and never sees more than the task in front of it. Privacy and the best capability, without trading one for the other. That's the future MLX is building toward.

Question 9

What is MLX?

Accepted Answer

The Apple Silicon Supercomputer: a community network that joins the Macs people already own into one AI computer. Apple chips have unified memory and fast neural engines, and hundreds of millions of them sit idle most of the day. Pooled together, they form a supercomputer owned by the people who power it. The MLX token records who builds the network and who does its work: Macs mine it for verified AI jobs. Every transaction is public, on a ledger whose history is anchored to Bitcoin and Solana.

Question 10

What is MLX Node?

Accepted Answer

A free Mac app with your MLX Wallet inside. It works for the network: it runs AI jobs when you allow it and mines MLX, hour by hour. MLX Node is the one download and keeps itself up to date. It also finds you the best AI for your Mac: the AI App Store is live on the web and in MLX Node, in a window of its own.

Question 11

What does a dedicated seat do today?

Accepted Answer

A dedicated seat reserves your Mac's place on the network, sets its share of the rewards and, as paid AI jobs arrive, puts it first in line for work. It pays nothing up front. While it runs on a Mac of its class that's linked to your wallet and doing AI work in MLX Node, it mines from the compute pool, hour by hour: up to 20× what a free seat mines for Core, 80× for Pro, 320× for Ultra, and 1.5× more for Latest. On a smaller Mac a seat mines at that Mac's class. Mined MLX is locked briefly: each day's mining unlocks after your seat's lock-in, or 30 days on a free seat.

Question 12

Why does a seat mine instead of paying out up front?

Accepted Answer

Because the network needs Macs that work, not seats that sit idle. Since 20 October 2026 a seat pays nothing when you buy it: it earns only while a Mac of its class does verified AI work. The halving curve still sets the network price of MLX, from one cent at genesis: what $100 buys halves, and the price doubles, once the first $20M has entered the network, then after $30M more, $40M more, and every $50M after that. A seat's dollars count at half on the curve.

Question 13

Is this a blockchain?

Accepted Answer

It's an open ledger. Every transaction is signed with your Ed25519 key, every block is hash-linked and signed by the sequencer, and all of it is public. You can verify any signature or block in your browser from the explorer. Every block also goes into a public transparency log whose checkpoints Macs running MLX Node check and cosign, anchored to Bitcoin and Solana, so history can't be rewritten without it showing. It runs as a single sequencer today so it can move fast while the community grows.

Question 14

How do lock-ins work?

Accepted Answer

Mined MLX is locked so the network grows with people who stay. With a dedicated seat, each day's mining unlocks after your seat's lock-in: 20 days for Nano, 14 for Lite, 10 for Core, 7 for Pro and 3 for Ultra, Latest editions the same, and the shortest seat in your wallet counts. On a free seat it's 30 days. The clock starts the day the MLX is mined. Your wallet shows locked and available MLX side by side, with the exact unlock date for each day of mining.

Question 15

Can I buy MLX without a dedicated seat?

Accepted Answer

Yes. Deposit any coin, then buy MLX with your USDC in the wallet. The protocol fills every order along the same curve: buying pays the curve price plus a premium and moves the price up, selling MLX back to USDC pays the curve price less a discount and moves it down. Spot MLX is yours right away, with no lock.

Question 16

How does the wallet work without a seed phrase?

Accepted Answer

When you create a wallet, your device makes a passkey for mlx.app, protected by Face ID or Touch ID and synced through iCloud Keychain. Your signing key comes from that passkey, so signing in on another Apple device just works. You can export a 24-word recovery phrase at any time as a backup. We never see your key.

Question 17

Which seat should I get?

Accepted Answer

We size seats for your Mac: Core for Macs with a base M-series chip, Pro for Pro and Max chips, Ultra for Ultra chips, and the Latest edition of each for Apple's newest chips (M6, M5 Pro, M5 Max, M5 Ultra). A seat mines in full on a Mac of its class; a bigger seat on a smaller Mac mines at that Mac's class. To start small, any Mac carries a Nano or Lite seat at its full rate.

Question 18

How do I pay for a dedicated seat?

Accepted Answer

With crypto, from any wallet or exchange: USDC or USDT on Solana, Ethereum, BNB Chain or Tron, or SOL, ETH, BNB, TRX and BTC. Each checkout gets a unique amount, so we can match your transfer without an account on another site. Your seat is minted on the MLX ledger as soon as the payment confirms, usually within a minute. You can also deposit first: any coin you send arrives as USDC on MLX Network, and a seat bought from that balance is recorded instantly.

The Apple Silicon Supercomputer.

Download

Terminal

Super easy.

The best AI for your Mac.

Your Mac, at work.

Why a network of Macs.

Why Apple Silicon. Why now.

The thesis

The network

Wallet and seats

Why Apple Silicon, and why now?

Built in the open.

The Apple Silicon Supercomputer.

Download

Terminal

Super easy.

01The App Store for AI.AI apps, tools and open models for the Mac and iPhone, each with what it costs, where your data goes and its licence, ranked by the MLX Score, made only from public data. Live on the web and in MLX Node.

The best AI for your Mac.

02Pick your seat.Buy once. A seat pays nothing up front: it lets your Mac mine up to 480× more MLX, hour by hour. Seat #22 is next.

03Or just hodl MLX.A wallet in seconds. Deposit any major coin, buy MLX at the live price, sell back any time.

04Seats mine. Hour by hour.A seat doesn't pay out up front. It sets how fast your Mac mines, and what it mines is locked for a few days, so the network grows with patient owners.

05Your Mac, at work.MLX Node runs AI work and earns MLX Tokens. A dedicated seat of your Mac's class multiplies it: 20× with Core, 80× with Pro, 320× with Ultra, and 1.5× more on Apple's newest chips.

Your Mac, at work.

06The network, live.Every node is public and verifiable.

07Why a network of Macs.GPU marketplaces have tried for years to rent out spare GPUs. Why would this work, and who uses the compute?

Why a network of Macs.

08Why Apple Silicon. Why now.The thesis, and the answers to every question we get, with sources.

Why Apple Silicon. Why now.

The thesis

The network

Wallet and seats

Why Apple Silicon, and why now?

09Built in the open.Explorer · Whitepaper · Builders · Bench · Integrity

Built in the open.

10Your seat is seconds away.It starts with MLX Node: it runs AI jobs on your Mac for the network and mines MLX Tokens.