The Innovation District - Salem River District

In Short

What we are proposing

An innovation district strategy for Salem: how to make this a place where physical AI companies form and hire, and where the people they hire can afford to live.

The chicken and egg

A company needs engineers it can hire before it will move here. An engineer needs more than one employer in town before they will move, because one job is one point of failure. Each side waits on the other.

Who moves first

Firms and individuals need a business case. Institutions can act without one, and in most of the districts we studied a university or a public body was in from the start.

What they set up

A neutral entity holding the ground. Test and fabrication space shared across firms. Housing inside the district rather than across town.

What it should produce

Companies forming and hiring here, and land values that rise across the whole area rather than one site. That is what pays for the phase after it (how that works).

How we would know

Whether the people already living in Salem end up better off. The recorded failure of these districts is homes that get expensive and push those people out. The study plan sets out the nine studies.

The Problem

Where the industry sits today

Oregon has the parts of this industry spread along the Willamette, and they are about a hundred miles apart. Companies do not relocate to a place without a labor market. Engineers do not move for a single job, because if it ends they need a comparable one nearby rather than a house to sell. And the technical people already here often work remotely for companies elsewhere, so their presence produces little local company formation. Neither side can move first. Meanwhile Oregon’s manufacturing employment was about 178,900 in 2025 and fell by roughly ten thousand in the year to June 2026 (Oregon Employment Department; the figures sit on Housing Supply).

Underneath that sits a slower leak. Oregon is good at the top of the funnel: it attracts technical people, educates them, and keeps about 64% of its university graduates employed in the state five years later: of 37,551 bachelor’s graduates in the 2013 to 2015 cohort, 24,000 were still working in Oregon after five years (Oregon Employment Department, November 2025). Then it loses them. An analysis of Census data released in January shows Oregon losing residents aged thirty to fifty faster than it gains them, with housing cost among the causes, while the people arriving earn considerably less than the people leaving (Oregon Journalism Project, March 2026). Because the state depends on income tax more than most, every departure is a revenue loss as well as a skills loss.

Why neither side moves first

Firms and individuals both need a business case before they move, and at the start neither has one. Institutions do not need a business case. In most of the districts we studied a public body or a university was among the founders: Stockholm and Ericsson set up a neutral foundation together in 1985, five St Louis anchors pooled about $29 million, Berlin physically relocated Humboldt’s science faculties onto the site. The cases.

An open market in a landscaped courtyard between reused industrial buildings — Salem, in the middle of the corridor, with standing industrial buildings and room for homes beside them. Concept render.

Definition

What is physical AI?

Software that senses the physical world, decides, and acts on it through hardware. A weeding robot, a warehouse handling system, a surgical assist arm and an autonomous shuttle are built by different industries, and underneath they are the same shape of problem.

The domain knowledge sits on top and differs in every column. The engineering underneath is largely the same, which is the argument for one district rather than five.

What different people call it

The name is unsettled. These all describe the same field.

Physical AI · investors and the press
Embodied intelligence, embodied AI · research groups
Robotics, controls, field robotics · most of the people who build it
Automation, autonomous systems · industry and operations
Applied/Physical AI · the Governor’s Prosperity Council, June 2026

We use physical AI on this page because it is the term most readers arrive with. Nothing here depends on which one you use.

The Proposition

Why an innovation district?

Several of the things that decide whether this work succeeds are easier to supply in one place than in five. These six come out of the districts we studied, and each one has a case behind it. The section after this one asks the same question about Salem.

Shared equipment nobody buys alone

Test rigs, fabrication and instrumentation are too expensive for one firm and too specific to rent. imec in Leuven was funded by the Flemish government as a neutral institute, and because no company owns it, rival chipmakers all fund and use it.

Cross-pollination between domains

A perception problem in an orchard and a perception problem in a warehouse are often the same problem. Kista's first act was one shared building putting university labs beside companies. Berlin moved Humboldt's science faculties onto the Adlershof site, and about 1,000 companies grew around them.

Companies form and grow here

Districts act as incubator and landlord at once, with small tenancies and room to fail cheaply. Cortex grew 406 companies on a St Louis brownfield. Cambridge Science Park seeded a cluster now spanning thousands of firms. Both took about two decades.

Talent gathers and stays

Engineers move between these domains more easily than the industries do, so a district lets someone change problem without changing city. Eindhoven's Brainport region holds roughly 73,000 technology jobs and about 40% of all Dutch research and development.

Capital and public tools arrive together

Five anchor institutions pooled about $29 million to start Cortex and paired it with tax increment financing. Barcelona used rezoning itself as the financing tool at 22@. Neither move is available to a single building.

Suppliers take your call

These products run in the thousands to tens of thousands a year. One firm at that volume is a nuisance to a component supplier. Several industries at that volume in one place is a customer worth retooling for.

What the districts also show, including the parts against us

Culture and community are treated in this literature as underwriting rather than decoration, and Medellín's Ruta N seated citizens as a full partner from the start. The honest counter-argument is that domains can diverge faster than they converge, shared facilities are harder to run than they look, and putting companies near each other can produce meetings rather than results. Cortex took two decades. The cases, and how six of them scored against seven patterns.

Homes inside a district are part of the original Brookings definition of the term, and the record shows them delivered unevenly: a minority built them in from the start, several added them decades later, and a few have none at all. Salem's version puts them in at the beginning, which is the subject of the river district.

A district street of reused buildings with workshops at ground level, people walking and small machines moving at walking pace — A street where the work can be tested in the open. Concept render.

Salem

Why an innovation district in Salem?

Oregon’s two public research universities are in Corvallis and Eugene, and the emerging innovation districts named in the state’s June 2026 Prosperity Council report are in Portland, in the southern part of the valley, and in eastern Oregon. Salem is a different kind of place, so the case for it rests on different things. These are the conditions here that bear on the question.

A factory for this technology is already here

An Oregon State spinout manufactures humanoid robots at RoboFab in Salem, reported as the first purpose-built factory of its kind. The research behind that technology is done in Corvallis. The manufacturing is done here.

Distance to the rest of the corridor

About 47 miles by road from Portland, 35 from Oregon State, 60 from the Hillsboro semiconductor cluster and 64 from Eugene. Salem is also the seat of state government.

Industrial buildings already standing

The west bank holds industrial buildings with high ceilings and heavy floors, which is what test and fabrication space needs. Adapting one is faster and cheaper than building new, and it lets a district open early and grow on real occupancy.

Land for homes beside the work

There is developable land inside the urban growth boundary, the line that sets how far the city can build outward. Salem’s median home price is lower than in the Portland metro, which is one of the places people move here from.

Training programs that already run

Chemeketa Community College runs robotics, industrial electronics and machining programs in Salem. Willamette University runs data science and dual-degree engineering. The technician end of the pipeline operates here today.

A large public and health employment base

State government accounts for about a quarter of the metro’s 191,200 nonfarm jobs. Salem Health employs about 6,400 people. Marion County holds 13 percent of Oregon’s food manufacturing employment.

How a Salem district would connect to university research, by partnering outward or by putting capacity here, is one of the questions in the study plan.

On how Salem compares with its peers, a stakeholder submission collected in the same Prosperity Council report cites the Brookings Metro Monitor for March 2026: among 85 US metro areas of between 250,000 and 500,000 people, Bend ranks first for prosperity and growth, Salem fourth for prosperity and seventh for growth, and Eugene twelfth for prosperity and thirtieth for growth. All three rank in the top fifth of all metro areas for inclusion.

None of the institutions or companies named on this page has been approached about this proposal, and none has any involvement in it.

Conditions

Why Oregon?

The parts are already here, spread along about a hundred miles of the Willamette between Portland and Eugene. Salem sits in the middle of them.

Semiconductors and hardware

The Hillsboro cluster, including Intel, Lattice, Qorvo, Thermo Fisher and Siltronic. Sensors, boards and precision assembly at scale.

Robotics

A research base at Oregon State, and a humanoid robot factory operating in Salem's Mill Creek employment area.

Agriculture

One of the most diverse farming valleys in the country, with about 220 commodities grown and 40,000 farms in the state.

Health and care

A regional hospital system in Salem, and an aging state population that makes the demand for care technology real.

Mass timber

An established sector, with the fabrication and material knowledge that off-site building depends on.

Ground to test on

Working farms, an interstate freight corridor, factories and clinics within a short drive of each other, so a system can be tried in the conditions it would actually run in.

These sit in different cities along the corridor. The corridor, company by company.

What the state has said about this

In June 2026 the Governor’s Prosperity Council listed Applied/Physical AI, including semiconductors and robotics, among the sectors it says Oregon is poised to grow and lead in. The same report recommends the state back catalytic public and private partnerships that can found emerging innovation districts, at investments on the order of $10 million per project, and gives three examples: the Confluence Innovation Campus in Portland, the Southern Willamette Valley Innovation Corridor, and industrial site development in eastern Oregon (Oregon Prosperity Council, June 2026).

Salem is not one of the three. We went looking for Salem elsewhere in the main report and did not find it. Those are recommendations to the Governor and the Legislature rather than money that has been appropriated, so the list is not closed.

In a letter to the same Council, OSU President Jayathi Murthy wrote: “As an anchor institution, OSU is prepared to support infrastructure and real estate development that enables economic growth. This includes innovation districts, research facilities, and industry-adjacent spaces.” That is a public statement of the university’s own intent. Nobody at Oregon State has been approached about anything on this page, and the university has no involvement in it. We quote it because five of the six districts in our scoring table put a neutral entity in charge of the mission, and in most of them a university or a public body was among the founders.

A mass timber panel line inside a reused building, with people working alongside the machinery — Mass timber fabrication, one of the sectors Oregon already has. Concept render.

The Approach

Where, when and how?

West Salem, in industrial buildings that are already standing, starting with one of them. Five moves, in order. The order is the part that matters: across ten turnarounds we studied, the places that worked spent cheap and fast first and drew heavy capital only after early wins proved demand. Sequence beats scale.

Convene

A champion institution, a university, the City, and employers in one room, agreeing what is worth testing.

Visit a working district together

Cortex, Brainport or the Brooklyn Navy Yard. The people who would build this should see one running before committing to it.

Form the holding entity

A neutral organization holding the ground and the tenancy decisions, with the university, the public sector, industry and the community represented.

Land the first anchor and the public tool

One tenant with a reason to be here, and a financing mechanism. In Salem that means urban renewal and tax increment.

Build the first proof

One adapted building, occupied, with the results published. Adaptive reuse comes first because it is faster and cheaper than new construction and because the buildings are already there.

Five of the six districts in our scoring table put a neutral entity fully in charge of the mission, and two of them, Kista and Newlab, began with a single shared building. How other districts were built. On how the ground itself creates value, District Value Creation.

Fabricators and sculptors working with the doors open in a reused warehouse bay — Existing industrial buildings on the west bank, adapted rather than replaced. Concept render.

Measurement

What outcomes would this be judged on?

Districts are usually measured by what gets built. The question that matters here is whether the people already living in Salem end up better off, so these are the figures we would publish, on a fixed schedule, whichever way they come out.

Companies still here in year five

Formed, moved in, or expanded, and still operating. Announcements are easy; survival is the test.

Engineers who chose to stay

People who took a job here over one elsewhere, and were still here three years later.

Median household income

For Salem as a whole, not for the district. A district that raises its own average and nothing else has failed.

Local business survival

Whether the businesses already trading nearby are still trading, and whether new ones are locally owned.

Homes, and who can afford them

Units delivered, and what share of them a household on a local wage can rent or buy.

Vehicle trips per household

Measured against the Salem average, which is the test of whether the place works as designed.

The full measurement set, drawn from Michael Shuman's twelve local-economy indicators, sits on District Value Creation.

An older resident, a caregiver and an assistive machine in a bright room — Every application on this page ends with a person. Concept render.

The Evidence

What this page is built on

Everything above draws on research we have published in full, including the parts that argue against us. Each page carries its sources.

Innovation District Precedents

Districts that were built, six of them scored against seven patterns, plus the partners behind each and how the buildings and money came together. Read it.

Oregon’s Physical AI Corridor

The companies, institutions and employment figures along the Willamette, company by company, with what we left out because we could not source it. Read it.

What Makes a District Work

Ten local economic development turnarounds, the ten patterns they share, what each cost, and how long each actually took. Read it.

How Districts Create Value

Seven comparable districts, where the value is actually made, who funds the infrastructure, and the twelve indicators we would be measured against. Read it.

The ULI Salem panel, 2022

An independent panel convened at the City of Salem’s request on this exact industrial area, and where our own premises differ from theirs. Read it.

Global District Developments

A running scan of the car-light, landscape-led districts cities and capital are committing to now, and what each one signals. Read it.

A shared kitchen and long table in a co-living block, with people cooking and eating together — The district would be a place to live as well as to work. Concept render.

In Closing

What we think, and what would settle it

Oregon has the parts of this industry and the ground to test it on, spread along about a hundred miles of the Willamette. Salem sits in the middle, has the buildings, and has room for the homes. That is the argument, and it rests on evidence from districts that were built, most of which took about twenty years.

What we do not know is whether the market supports it, whether the ground and its infrastructure can carry it, what the public sector would actually do, and how many jobs it would produce. Those are the market study, the infrastructure planning study, the public private partnership work and the job creation study in the study plan, and they should be commissioned by a public or institutional sponsor rather than by us, then published in full whichever way they come out.

Three of the four are treated here, and the public private partnership work is taken up under Where, when and how. The market study would establish what a first phase can be financed against today, and separately what would have to change to support more. The infrastructure planning study would sort every finding into no material issue, solvable at a known cost, or a possible fatal flaw. The job creation study would produce a number with its method published beside it, because a figure produced before the study is a figure produced by whoever wants it to be large.

If you build this technology for a living and think part of the argument is wrong, that is the most useful thing you could tell us. [email protected]