It’s all just a little bit of history repeating itself…
Bottom Line Up Front
Three days ago, Swiss researchers eliminated the need for giant AI data centers for 80-90% of what people actually use AI for. Most people haven’t noticed yet. But this is the moment, like when email went from ‘something technical people use’ to ‘how everyone communicates’, except it’s happening in months, not years.
When technology hits the “good enough AND cheaper” threshold, change happens FAST. Distributed computing didn’t gradually replace mainframes over 20 years—it happened in 3-5 years once the technology crossed that line.
We just crossed that line. Last week.
This matters for three reasons workers and communities should care about:
First: The job opportunities this creates don’t require billion-dollar facilities or taxpayer subsidies. They happen in existing offices, across every industry, creating distributed implementation and training jobs instead of centralized data center positions.
Second: There’s a massive market gap nobody’s talking about. The “missing middle” between giant data centers and do-it-yourself local AI. Local AI hosting for small businesses and individuals could create thousands of small businesses across Michigan, each employing 3-10 people, with no tax breaks required.
Third: The Michigan data centers being approved right now – Saline’s $7 billion, Van Buren’s 1 gigawatt, Lyon’s 1.8 million square feet – are betting on a model that might serve 20% of the market, not 80%, by the time they finish construction in 2027-2028.
Communities are subsidizing yesterday’s infrastructure with 25-year tax breaks just as the technology fundamentally shifts.
This article follows up on my prior piece – What the 1980s-90s IT Revolution Teaches Us About AI. History does, in fact, repeat itself.
*Note: I reference Michigan in the following. The same is true of ANY state. Most of my tracking right now is in Michigan.
What Just Happened: Where AI Actually Runs
Most People Don’t Understand This (But It Matters A LOT)
If you want to understand the basic basics of how this works, this section is for you.
Using AI like Claude.ai Website or the Claude App:
What happens when you type a question:
- You type on your computer/phone
- Your request travels through the internet to Anthropic’s data center (could be anywhere – California, Virginia, etc.)
- Giant computers in that warehouse process your question
- Answer travels back through the internet to you
- Your data lives on Anthropic’s servers, not your computer
Why this matters:
- Your question left your home or building
- Someone else’s computers processed it
- Your data is stored somewhere else
- Requires internet connection
- You or your company have no control over where data goes
Cloud/Internet/Server Farm (Same Idea, Different Companies):
Examples: Google Workspace, Microsoft 365, Salesforce, AWS
What “the cloud” actually means:
- “The cloud” = someone else’s giant warehouse of computers
- Could be Amazon’s, Microsoft’s, Google’s data centers
- Your work happens on their computers, not yours
- Everything travels through the internet to get there
Think of it like this:
- Before: You kept files in your office filing cabinet
- Cloud: You ship files to a storage facility across town to file them
- Every time you need a file, they ship it back
- Convenient! But you don’t control the warehouse
- Paid service
Local Deployment in Your Company (Enterprise):
What Anyway Systems and similar technologies enable:
What happens with local AI:
- You download the AI software once (like downloading Microsoft Word)
- It runs on 4-10 computers in YOUR office
- Questions get answered by YOUR computers
- Data never leaves your building
- No internet required once it’s set up
Why this is revolutionary:
- Your data stays in your building
- Your computers do the work
- No monthly cloud bills
- No dependence on someone else’s data center
- You control everything
The Analogy:
| Model | Real-World Comparison |
|---|---|
| Cloud/Server Farm | Mailing your tax documents to H&R Block in another city, waiting for them to process, mailing back |
| Local Deployment | Having a tax preparer come to your office with their laptop and do it right there |
Both get the job done. But one keeps everything in your building.
The Missing Middle: Who Serves Everyone Else?
Here’s an Idea I Don’t Know that Anyone is Talking About
Anyway Systems proves big companies can run AI on 4-10 computers in their office. Great for enterprises and some small businesses.
But what about:
- The corner law firm with 5 employees?
- The accountant who wants AI but has one laptop?
- The small manufacturer with 15 workers and confidential designs?
- The medical practice with patient privacy requirements?
- Individual professionals who need AI but not Big Tech looking at their data?
They can’t afford 4-10 dedicated computers. But they also don’t want their data in Amazon’s cloud 500 miles away.
Enter: Local AI Service Providers
Think of them like:
- 1990s Internet Service Providers (your friendly local internet company)
- 2000s Managed Service Providers (your local IT support)
- 2026: Local AI Service Provider (your local AI computing). LASP, if you will. I think I created a new acronym.
The Business Model:
What they offer:
- Small facility (2,000-5,000 sq ft) in your community
- 10-50 computers with AI capabilities
- Serve 50-200 local small businesses and individuals
- Monthly subscription: $50-200/month per client
- Your data stays LOCAL (same city/county)
- Call someone who knows your name
- Privacy-focused alternative to Big Tech
The Infrastructure Needed:
Not a hyperscale data center:
- 2,000-5,000 sq ft (like a small warehouse or large office)
- 10-50 computers with GPUs
- Standard commercial space (industrial park, downtown building, converted retail)
- Normal power requirements (not gigawatts)
- Standard internet connection
Investment Required:
- $500,000 – $2 million to start
- Uses existing commercial space
- Normal small business financing
- Equipment purchased over time
- Can start small and scale
Compare to Hyperscale:
- Saline: 2.2 million sq ft, 1.4 gigawatts, $7 billion, 25-year tax breaks
- Local AI Hosting: 5,000 sq ft, normal power, $500K-2M, no tax breaks needed
The Job Creation Math
One Local AI Service Provider:
- 3-10 employees (tech staff, support, sales, management)
- Serves 50-200 clients
- Annual revenue: $300K-$2M (depending on client count)
- Normal tax-paying small business
If Michigan had 50 Local AI Service Providers:
- 150-500 jobs created
- Distributed across 50 communities (not 3 locations)
- Local ownership possible (not Oracle/Microsoft/Google)
- Tax-paying businesses generating revenue for communities
- Services stay affordable for local small businesses
Compare to Three Hyperscale Data Centers:
- Saline: 45 permanent jobs promised
- Van Buren: 30-50 estimated
- Lyon: 50-100 estimated
- Total: ~135 jobs (being generous)
- 3 locations only
- Out-of-state corporate ownership
- 25-year tax exemptions (no revenue for communities)
The math is stunning:
- 50 small businesses create 3x more jobs than 3 hyperscale facilities
- Jobs distributed across entire state instead of 3 townships
- No taxpayer subsidies required
- Communities collect taxes instead of giving them away
- Local ownership keeps wealth in Michigan
The 1990s Parallel
What happened with Internet Service Providers:
1994-1998: The Boom
- Thousands of local ISPs emerged across America
- Names like “Your Town Internet” or “Local Web Services”
- Served neighborhoods, small towns Big Tech ignored
- Personal service, local support, knew your business
1998-2005: Consolidation
- AOL, Comcast, and others bought many local ISPs
- But hundreds of local/regional ISPs survived and thrived
- Especially in rural areas and underserved markets
2026: Today
- Thousands of independent ISPs still operate
- They serve niches Big Tech doesn’t want
- Local, personal service still matters
- Many are highly profitable small businesses
AI hosting will follow the same pattern:
- 2026-2028: Local AI providers emerge everywhere
- Serve niches Big Tech ignores (privacy-focused, local support, reasonable pricing)
- Some consolidation but thousands will survive
- This is the window to be early
Examples Already Emerging
From the industry:
- Akamai advertising edge-based AI (bringing compute closer to users)
- Cisco promoting Unified Edge systems (local + cloud hybrid)
- “Neoclouds” growing 82% annually (smaller regional AI hosting)
But these are still corporate. The LOCAL, small-business-focused providers haven’t emerged yet.
That’s the opportunity.
Who This Serves
Small Businesses:
- 5-50 employees can’t afford dedicated AI infrastructure
- Need AI for customer service, document processing, data analysis
- Want data privacy (client confidential information)
- Value local support (call someone who answers)
- Can’t justify $100K+ investment in equipment
Individual Professionals:
- Lawyers with client confidentiality requirements
- Accountants processing sensitive financial data
- Doctors with patient information (HIPAA compliance)
- Consultants who want AI but not Big Tech surveillance
- Anyone privacy-conscious
Rural Communities:
- Poor internet makes cloud AI frustrating
- Local hosting provides better performance
- Underserved by Big Tech
- Keep money in local economy
The Business Opportunity
Starting a Local AI Service Provider requires:
Technical Skills:
- IT background (networking, systems administration)
- Understanding of AI tools and models
- Or partner with someone who has these skills
Business Skills:
- Customer service focus
- Marketing to local small businesses
- Financial management
- Or partner with someone with business experience
Investment:
- Start with $500K (small operation, 10-20 clients)
- Scale to $2M (50-100 clients, more equipment)
- Can use SBA loans, local bank financing
- Equipment can be purchased incrementally
Revenue Model:
- Monthly subscriptions: $50-200/client
- 50 clients × $100/month = $60K annual revenue
- 100 clients × $150/month = $180K annual revenue
- Scale from there
Compare to competing for data center jobs:
- Data center: 30-50 permanent positions total
- Need to be in the right location (Saline/Van Buren/Lyon)
- Jobs controlled by out-of-state corporations
- No ownership opportunity
vs.
- Local AI hosting: Create your own business
- 50 potential providers across Michigan
- Local ownership keeps wealth in state
- Expandable as market grows
[We’ll provide detailed step-by-step guidance in Under the Radar this Friday, including technical requirements, business planning, financing options, and how to get started.]
What Still Needs Giant Facilities?
Let’s be clear: Hyperscale data centers won’t disappear. But their role is changing.
Three things still need massive facilities:
1. Training Brand-New AI Models from Scratch
- Creating the next ChatGPT or Claude from zero
- Months of computing time
- Billions of dollars in equipment
- Very few companies do this (OpenAI, Google, Anthropic, Meta, maybe 10-20 worldwide)
2. Serving Hundreds of Millions of Users Simultaneously
- When 100+ million people use ChatGPT at the exact same time
- Global-scale consumer products
- Requires massive coordination infrastructure
- That’s what Big Tech’s data centers do well
3. Cutting-Edge AI Research
- Experimental models that don’t exist yet
- University research programs
- Corporate R&D for breakthrough technologies
But here’s what MOST businesses and individuals actually need:
- Customer service AI answering routine questions
- Document analysis and processing
- Data analysis and reporting
- Workflow automation and productivity tools
- Department-specific applications
- Help with routine tasks
ALL of that can now run either:
- On a few computers in your office (if you’re an enterprise), or
- Through a Local AI Service Provider (if you’re a small business or individual)
The market split:
- Maybe 20% of AI work still needs hyperscale (training new models, global consumer services)
- The other 80% can be distributed
The timeline to “complete obsolescence” of the current model:
- Hyperscale data centers won’t become obsolete
- But they’ll serve 20% of the market instead of 80%
- That shift happens 2026-2030
- Exactly when today’s approved data centers will be opening
What this means for Saline/Van Buren/Lyon:
- They’re being built for the OLD model (80% of AI work needs hyperscale)
- They’ll open into the NEW model (maybe 20% needs hyperscale)
- Construction finishes 2027-2028 just as the market fundamentally shifts
Why This Creates Different Jobs
Three Models, Three Job Creation Patterns
| Hyperscale Data Centers | Local AI Service Providers | AI Implementation Consulting | |
|---|---|---|---|
| Investment | $7 billion (Saline) | $500K-$2M per provider | Minimal (laptop, skills, network) |
| Facility Size | 2.2 million sq ft | 2,000-5,000 sq ft | Your home office or shared workspace |
| Jobs Per Location | 30-50 permanent | 3-10 employees | 1-5 person consultancy |
| Geographic Distribution | 3 facilities statewide | 50+ across state | Thousands statewide |
| Total Jobs Created | ~135 in Michigan | 150-500 in Michigan | 2,000-5,000+ in Michigan |
| Job Duration | Permanent (if facility stays open) | Permanent small businesses | Ongoing consulting (decades) |
| Construction Jobs | 2,500 for 18-24 months | Minimal (renovate existing space) | None (uses existing offices) |
| Ownership | Out-of-state corporations | Local ownership possible | Local individuals/small firms |
| Taxpayer Subsidy | 25-year tax exemptions (~$420M for Saline) | None needed | None needed |
| Tax Revenue Generated | $0 for 25 years | Normal business taxes from day one | Normal business taxes from day one |
| Barrier to Entry | Impossible (need billions) | Achievable ($500K-2M, SBA loans) | Low (education, skills, networking) |
| Benefits | Tech companies | Every industry | Every industry |
The pattern is clear:
The distributed models (Local AI Service Providers + Implementation Consulting) create:
- 10-40x more jobs than hyperscale data centers
- Jobs accessible to Michigan workers and entrepreneurs
- No taxpayer subsidies required
- Tax revenue from day one instead of 25-year exemptions
- Local ownership opportunities keeping wealth in state
- Distributed across communities of all sizes
The Implementation Wave: Jobs That Start Tomorrow
AI Implementation Specialist
What they do:
- Help businesses set up local AI systems (enterprise deployments)
- Connect AI to existing software and workflows
- Customize AI for specific business needs
- Train employees on effective AI use
Where jobs exist:
- Every industry (manufacturing, healthcare, legal, finance, retail)
- Companies of all sizes
- Distributed across entire economy
Skills needed:
- Understanding business processes (how work actually gets done)
- Problem-solving (where does AI make sense?)
- Communication (explaining tech to non-tech people)
- Some technical background helpful but not required
AI Trainer
What they do:
- Teach employees how to use AI tools effectively
- Develop best practices for specific industries
- Quality control and results measurement
- Ongoing optimization and improvement
Skills needed:
- Training and teaching ability
- Industry knowledge
- Communication skills
- Understanding of AI capabilities and limitations
Multi-Agent Coordinator
What they do:
- Set up AI “teams” that work together
- Like a project manager for AI assistants
- Ensure different AI systems communicate and coordinate
- Optimize workflows across AI agents
Skills needed:
- Project management experience
- Systems thinking
- Process design
- Technical aptitude
Use Case Developer
What they do:
- Figure out WHERE AI helps specific businesses
- Design workflows that incorporate AI effectively
- Measure results and ROI
- Identify new opportunities as technology evolves
Skills needed:
- Business analysis
- Creative problem-solving
- Industry knowledge
- Data analysis
Local AI Service Provider Owner/Operator
What they do:
- Run local AI hosting business
- Manage technical infrastructure
- Provide customer support
- Market to local businesses
- Scale operations as demand grows
Skills needed:
- IT/systems administration (or hire someone who has it)
- Business management
- Customer service
- Sales and marketing
The Timeline: Are We Being Dramatic?
No. Here’s why this moment matters.
The Window is NOW
2026 (RIGHT NOW – January):
- Technology works and is available (Anyway Systems launched 3 days ago)
- 57% of companies already deploying AI agents in production
- Qualcomm launching full robotics suite today (Monday, January 5)
- Most people haven’t heard about local AI yet
- Early movers get 2-3 year advantage
- Local AI Service Provider market is WIDE OPEN
2027:
- Standard practice for forward-thinking companies
- First wave of Local AI Service Providers established
- Like having a company website in 1998
- Competitors start asking “why don’t we have this?”
2028:
- Expected baseline capability
- 50+ Local AI Service Providers operating in your state
- Like having email in 2005
- Companies WITHOUT local AI look behind
- First hyperscale data centers opening (Saline, Van Buren if approved)
2030:
- Universal standard
- Like having smartphones in 2015
- Can’t imagine working without it
- Distributed model is normal
- Hyperscale serves niche market (training new models, global consumer services)
The 1995 Internet Parallel
Adoption Speed:
- 1995: 16% of Americans used internet
- 1998: 36%
- 2000: 52%
- Three-five year window from “early adopter” to “standard practice”
What happened to those who moved early:
- Started local ISPs: Many sold for millions, some still operate profitably
- Learned web development: Decades of high-paying work
- Built e-commerce sites: First-mover advantage worth billions
- Started digital marketing agencies: Built sustainable businesses
What happened to those who waited:
- Scrambled to catch up in 2000-2005
- Paid consultants to do what they could have learned
- Lost first-mover advantage
- Competed in crowded markets
We’re in 1995 right now. The infrastructure just became “good enough.” The race is on.
Where to Start
Good news: You don’t need to be a programmer or have millions to invest
For AI Implementation/Consulting Careers:
What matters most:
- Understanding business processes (how work actually gets done in your industry)
- Problem-solving skills (where does AI make sense?)
- Communication ability (explaining tech to non-tech people)
- Project management experience
- Industry knowledge (manufacturing, healthcare, legal, whatever you know)
The 1990s parallel:
- Best networking consultants weren’t always the most technical
- They understood BOTH technology AND business
- They could translate between the two
- Same will be true for AI implementation
For Starting a Local AI Service Provider:
What you need:
- Technical: IT background, systems administration, or partner with someone who has these skills
- Business: Customer service focus, local marketing, financial management
- Capital: $500K to start small (10-20 clients), scale to $2M (50-100 clients)
- Space: 2,000-5,000 sq ft commercial (industrial park, office building, converted retail)
- Financing: SBA loans, local bank financing, angel investors
Resources and detailed guidance: We’ll provide step-by-step instructions in Under the Radar this Friday (January 9), including:
- Technical requirements (what equipment, what software, what setup)
- Business planning (how to find clients, what to charge, how to market)
- Financing options (SBA loans, local banks, investor pitch templates)
- Legal setup (business structure, liability, contracts)
- Real cost breakdowns (equipment, space, labor, operations)
- First 100 days roadmap (how to launch and scale)
Plus: Discord community for Michigan workers exploring these opportunities, sharing resources, and supporting each other through the transition.
What This Means for Michigan
For Workers:
This is your window to position yourself for the implementation wave. Not scramble when everyone realizes they need help.
The opportunities:
- AI Implementation Consulting: Start tomorrow with skills you have
- Local AI Service Provider: Start in 6-12 months with proper planning
- Both: Jobs distributed across EVERY community, not just data center towns
The timing:
- 2026: Learn, position, launch
- 2027: Establish, grow, expand
- 2028: Competitive advantage, market leadership
For Entrepreneurs:
The Local AI Service Provider market is wide open RIGHT NOW.
Consider:
- 50+ providers could operate profitably in Michigan
- Current number: Approximately zero
- Investment: $500K-2M (achievable with SBA loans)
- Market demand: Thousands of small businesses need this
- Competition: Minimal (you’re early)
This is like being positioned to start a local ISP in 1994 or a web development agency in 1996.
First movers will have 2-3 years before competition catches up.
For Communities:
The data centers being approved today are betting billions on yesterday’s model:
Saline:
- Investment: $7 billion
- Construction starts: 2026
- Opens: 2028
- Tax exemption: Through 2050 (25 years)
- Jobs promised: 450 permanent
- Cost per job: $933,000 in tax subsidies
Van Buren:
- Decision: January 14, 2026 (next week)
- Same pattern as Saline
- 25-year tax exemption if approved
Lyon:
- Already approved
- Phased construction through 2031
- Residents only learned about it in December (approved in September without public hearings)
By the time these facilities are fully operational (2027-2031):
- Distributed AI will be standard practice
- Hyperscale centers will serve 20% of market (training, global services)
- The 80% of work that creates most jobs will be distributed
- Communities will be 5-25 years into tax exemptions with no revenue
Meanwhile, the REAL job creation (Local AI Service Providers + Implementation Consulting):
- Creates 10-40x more jobs
- Distributed across every community
- Requires NO taxpayer subsidies
- Generates tax revenue from day one
- Enables local ownership
- Serves local businesses that need AI
Communities are giving away tax revenue through 2050 for centralized facilities that might serve a shrinking market, while the actual job creation needs no subsidies at all.
The Critical Juncture
The pattern from previous tech shifts is clear: When technology crosses the “good enough AND cheaper” threshold, adoption accelerates exponentially.
Distributed computing didn’t gradually replace mainframes. Email didn’t slowly replace memos. Smartphones didn’t cautiously replace flip phones.
The shift happened in 3-5 years once the technology was ready. This technology? Moves much, much faster.
AI Just Got Ready. Last Week.
January 2, 2026: Swiss researchers (Anyway Systems) proved that 80-90% of AI work doesn’t need giant data centers. It can run on 4-10 computers in an office – or through a local AI service provider serving your community.
January 5, 2026 (today): Qualcomm announces full robotics suite. More distributed AI applications.
January 14, 2026 (next week): Van Buren Township votes on 1-gigawatt data center. Communities still approving yesterday’s infrastructure.
Three Paths Forward
Path 1: The Worker/Consultant
- Learn AI implementation skills NOW
- Position yourself for implementation wave
- Start consulting in 2026-2027
- Decades of high-demand work ahead
- Minimal investment, maximum opportunity
Path 2: The Entrepreneur
- Start Local AI Service Provider in 2026-2027
- Serve underserved small business market
- $500K-2M investment (achievable with financing)
- First-mover advantage worth millions
- Build valuable, sellable business
Path 3: The Wait-and-See
- Keep hoping data center jobs materialize
- Wait until 2028-2030 to understand shift
- Compete in crowded market with prepared competitors
- Scramble to catch up
- Pay others to do what you could have learned
The Question
The question isn’t whether this transformation will happen.
The technology is ready. The market is moving (57% in production already). The investment is pouring in ($7.8 billion in 2025 growing to $52 billion by 2030). The job categories are forming.
The question is whether you’ll be positioned to benefit from it:
- As a worker who can help businesses deploy distributed AI
- As an entrepreneur who starts a Local AI Service Provider
- As a community that recognizes where the actual job creation will occur
This is 1995. The infrastructure just became viable.
What you do in the next 12 months determines whether you’re ahead of the wave or scrambling to catch up.
Next Steps
This Friday (January 9) in Under the Radar:
- Detailed guide to starting a Local AI Service Provider
- Technical requirements and equipment specs
- Business planning and financial projections
- Where to get training for AI implementation consulting
- How to connect with others exploring these opportunities
- Discord community for Michigan workers in transition
This Sunday (January 11) in PivotIntel Weekly:
- Qualcomm robotics announcement analysis
- Van Buren Township vote reminder (January 14)
- Latest on Saline construction
- Data center financial stability tracking
- Employment data updates
The inflection point is here. The window is open. The question is: What will you do with it?
Published by The Open Record L3C | Monday, January 5, 2026
Sources: Anyway Systems (EPFL), New Atlas, Gartner, Forrester, Deloitte, IDC, G2, VentureBeat, Messari State of AI 2025, Cloud Wars, SDxCentral, multiple industry analysts. All sources archived via Wayback Machine.
