Preventing Xometry 2.0

A shop owner sees the trap before the first setup is scheduled. The buyer is hidden, the platform owns the work order, the price arrives pre-shaped, and the shop is told where communication may occur. The objection is blunt: this is Xometry 2.0, another rent-seeking middleman turning American suppliers into anonymous capacity. A smart operator believes it because the model already exists in public view, and the public record shows the pattern — instant quoting, global supplier matching, relationship control, and contractual penalties around direct buyer contact.

I build in this category, so the objection lands on me too. It should. It is the right thing to demand of anyone who puts himself between a buyer and a shop.

The promise of this series is that American manufacturing needs coordination without surrender. Shops need cleaner demand, better routing, stronger verification, and less wasted quoting. The claim of this paper is falsifiable: a manufacturing coordination system becomes Xometry 2.0 when it controls price, blocks direct buyer-supplier relationships, monetizes transaction spread, or sells routing power; it avoids that failure only when suppliers set their own prices, buyers know who they are dealing with, routing rules are auditable, and platform revenue does not depend on capturing margin from the purchase order.

The category error is calling every manufacturing platform "coordination." Coordination improves the RFQ before it reaches the supplier. Rent-seeking captures the transaction after the buyer already wants the part. The difference shows up in four places: who sets the quote, who owns the customer relationship, who controls the message thread, and who benefits when the supplier's margin gets squeezed.

The critic's best case

The critic's best case starts with the customer record. In custom manufacturing, the buyer relationship is not a side benefit. It is the asset. A supplier learns whether the buyer sends clean drawings, pays on time, answers clarification requests, honors revisions, and comes back with serious work. When a platform captures that record, the shop keeps the operational burden while the intermediary keeps the commercial memory.

Xometry is the obvious case because it says the quiet part loudly enough. Its own materials describe a marketplace where buyers can upload engineering schematics, receive instant pricing and lead times, and connect to a global supplier network. Its 2024 annual report described 4,375 active suppliers and an AI-driven quoting engine that provides instant pricing, lead times, and manufacturability feedback. That is a controlled transaction environment built around the drawing package, not a neutral bulletin board.

The relationship boundary is even clearer in Xometry's supplier terms. Partners are told to communicate through WorkCenter, follow anonymity and non-solicitation rules, avoid direct marketing or direct customer contact outside approved channels, and face a prohibited communication fee when they violate those rules. Xometry's own terms describe that fee as reimbursement for the burden and cost of developing the customer relationship.

Read that phrase twice, because I think it is the most honest sentence in the category. It says the platform considers the customer relationship its property. The shop may machine the part, carry the schedule risk, answer the quality questions, and ship the order, but the account belongs to the intermediary. The supplier is left with a job. The platform keeps the demand source.

The broader rent-seeker map

Xometry is the cleanest target, but it is not the whole pattern. The manufacturing platform world is filling up with companies promising instant access, managed sourcing, global networks, supplier optimization, AI matching, and one-click RFQs. I'll grant that some of it is genuinely useful. Some of it is a toll booth with better copywriting.

Fictiv belongs in the frame because it openly presents itself as a digital manufacturing and supply chain platform with instant quoting, auto DFM tools, global and domestic options, and complete order visibility. Its 2025 production manufacturing announcement described four global manufacturing centers in India, Mexico, China, and the United States. That is not automatically bad, but it is a different thesis from rebuilding enforceable domestic supplier relationships.

Zetwerk belongs in the frame because it presents itself as a global manufacturing network built around cost, lead time, supplier selection, tracking, and quality control. Its own site emphasizes globally competitive costs, lead time reduction, and large-scale production capacity. Zetwerk can obviously move parts; the open question is whether global network optimization strengthens American industrial depth or quietly turns domestic sourcing into one option inside a cost-arbitrage router.

Thomasnet belongs in the frame because Xometry bought it. Xometry announced the acquisition of Thomas in 2021 for $300 million, combining an on-demand manufacturing marketplace with a supplier discovery and digital marketing platform. Thomas later described itself as connecting buyers with more than 500,000 North American suppliers as part of Xometry's broader marketplace and cloud services stack. That matters because supplier discovery, advertising, quoting, and marketplace routing can collapse into one power center.

Vendra is the interesting edge case. Its YC page says it helps buyers submit one RFQ, automates sourcing, matches buyers with the right suppliers, handles outreach, collects competitive quotes, and claims "full transparency and no middleman markups." Its YC launch page says buyers collaborate directly with suppliers and choose exactly who builds the parts. Those claims put Vendra on the right side of the line if they stay true in practice. The test is whether its routing stays transparent, suppliers own their quotes, and successful work can turn into direct relationship value.

That is the point. The villain is not software. The villain is the toll-booth model. The same failure can come from an instant-quote broker, a managed marketplace, a global sourcing agent, a paid supplier directory, a procurement portal, or an AI matching tool. Once the platform controls price, hides the relationship, sells placement, or taxes every PO, it has crossed the line. That is the company I refuse to build.

How the bad version happens

The bad version begins when the platform sets the price. A supplier quote is a business judgment. It prices uncertainty, capacity, quality risk, delivery pressure, and the opportunity cost of taking the job at all. When the platform attaches the number first, the supplier is pushed into an accept-or-decline posture and no longer owns the quote as a commercial decision. What arrives is a work order shaped by someone else's margin target.

The next failure is relationship capture. A buyer and supplier who complete good work together should get more efficient. The second order should take less friction than the first, and the third should carry more trust than the second. A platform that treats repeat direct work as leakage has the wrong incentive baked in at the root. It has to police the message thread, because its revenue depends on keeping the buyer and the supplier apart.

Platform economics explains why this keeps happening. Two-sided marketplace research describes platforms that match buyers and sellers while extracting commissions from completed transactions, and it identifies disintermediation as a threat when buyers and sellers bypass the platform to avoid fees. That is exactly the pressure in custom manufacturing once a buyer trusts a supplier. The platform then faces a choice: add enough value to stay relevant, or enforce dependency through control.

The third failure is paid visibility. Supplier discovery looks harmless right up until ranking becomes a product. If placement can be bought, the buyer is no longer seeing the best fit; the buyer is seeing whoever paid for proximity to demand. In a low-risk catalog purchase that is annoying. In custom manufacturing it corrupts source selection, because what the buyer needs is fit, quality discipline, and delivery reliability, not the loudest bidder for attention.

Fourth comes global arbitrage dressed up as resilience. Global supply chains can be necessary, but they are not neutral. If a platform routes American demand through a global cost optimizer, domestic suppliers are forced to compete against labor, energy, regulatory, and enforcement regimes that are not comparable to their own. A U.S. shop carrying American compliance costs cannot rebuild capacity by accepting overseas price expectations filtered through a platform margin model.

Fifth is IP concentration. A custom RFQ package contains far more than a part request. It can reveal product strategy, engineering intent, production timing, and supplier priorities. A platform that centralizes drawings, models, quotes, supplier responses, buyer behavior, and award history becomes a high-value industrial intelligence repository. Theft is the obvious worry and the smaller one. The bigger danger is that sensitive demand signals become platform assets before they ever become supplier relationships.

The governed alternative

A governed coordination layer draws the boundary at control. It may structure the RFQ, verify supplier identity, route demand, record outcomes, and preserve accountability. It may not set supplier prices, hide the supplier from the buyer, sell ranking as fit, or punish a legitimate direct relationship after successful work. The purchase order is the enforcement boundary.

The main mechanism is revenue neutrality at the transaction. A platform that earns a hidden spread from the PO will eventually optimize for spread. A platform that earns through workflow, verification, governance, enterprise controls, or supplier tools can afford to let relationships compound. That single distinction decides whether the system strengthens the industrial base or farms it.

The routing rule has to be auditable. A buyer should know why a given supplier appeared. A supplier should know which evidence improves its future visibility. Route on verified capability, response behavior, delivery history, location, certification, or buyer preference and the system can defend itself in daylight. Route on paid placement, margin potential, or platform-owned supply balancing and the system is steering under cover of neutrality.

The abuse paths are easy to name. A supplier can exaggerate capability to get into better RFQs. A buyer can spam quote requests with no intent to award. A platform can bias routing toward revenue. A paid directory can sell visibility as if it were trust. A global network can call cost arbitrage "capacity." A governed system answers with verified records, buyer reputation, quote discipline, outcome memory, placement disclosure, and penalties for spam, quality drift, or false capability claims.

The failure signals are measurable, and any honest operator should publish them. Supplier churn rising while RFQ volume rises means the platform is burning supply. Repeat award rate falling means the system is creating transactions without trust. If direct-contact penalties are climbing, the platform is defending dependency. If supplier margins are falling while disputes rise, the marketplace is buying price collapse. And when paid placement outperforms verified performance, routing has quietly become advertising.

Implications

If the critic is right, Xometry 2.0 will look busy long before it looks destructive. Buyers will see fast quotes. Shops will see more work orders. Investors will see transaction volume. Then the real signals show up: weak repeat relationships, compressed supplier economics, more supplier churn, more leakage attempts, more routing opacity, and more platform control over who is allowed to talk to whom.

If the governed model works, the data should move the other way. RFQs get cleaner. Quote response time improves. Repeat awards rise. Disputes fall. Suppliers retain pricing power. Buyers build stronger direct supplier memory. The platform should be most valuable before source selection and less necessary after trust has formed.

American manufacturing does not need another extraction layer dressed up as digitization. It needs coordination that helps real buyers find real suppliers, then lets performance become durable trust. Xometry is the warning label, Fictiv and Zetwerk show the global-network version, Thomasnet shows how discovery can be absorbed into marketplace power, and Vendra shows the new entrant test: transparency claims only matter if the control points prove them.

The practical failure mode is rent extraction. The platform becomes more important than the supplier, the buyer, the quote, the PO, and the delivered part. That is how manufacturing software becomes industrial cancer — it feeds on the relationship tissue the country needs to rebuild. The next objection follows naturally, and it points at me just as directly: even if a coordination layer avoids Xometry 2.0, what stops visibility from becoming pay-to-play?

Questions to Ask

1. Who sets the final quote price?
2. Can the buyer see and choose the actual supplier?
3. Can successful work turn into a direct repeat relationship?
4. Does the platform make money from transaction spread, paid placement, workflow value, or verification?
5. Are routing rules visible enough to audit?
6. What metric would prove the platform is drifting into toll-booth behavior: falling repeat awards, supplier churn, margin compression, direct-contact penalties, paid-placement dominance, or rising dispute rates?