PivotIntel Weekly Intelligence Report #7

January 4, 2026

Bottom Line Up Front

Swiss researchers just proved you can run powerful AI models like ChatGPT on 4 computers instead of billion-dollar data centers. Michigan is approving 25-year tax breaks for $7 billion hyperscale facilities that may become obsolete before the subsidies expire. Meanwhile, bipartisan legislation to repeal all data center tax incentives gained momentum, Oracle broke ground in Saline despite equity partner withdrawal, and Van Buren Township votes January 14 on a 1-gigawatt facility. Communities have 11 days to leverage this intelligence.

The technology that could make hyperscale data centers unnecessary exists right now. Communities negotiating with developers should be asking: “Why do you need 575 acres and 1.4 gigawatts when Swiss researchers run the same AI on 4 computers?”

Anyway Systems Timeline: Not Coming Soon – HERE NOW

The acceleration pattern says this isn’t future technology. It’s deploying faster than data centers can be built.

Enterprise/Organizational Market: Q1-Q2 2026 (Already Launching)

Evidence this is past beta stage:

✅ Commercial company – EPFL researchers spun off Anyway Systems (anyway.dev), not research project
✅ Selling now – “software they’re now selling through their own company” (not “will sell”)
✅ Production-ready – 30-minute installation time signals finished product, not prototype
✅ Proven deployment – ChatGPT-120B running on 4 computers is proof-of-concept AND proof-of-deployment

Target customers deploying NOW:

European organizations (GDPR compliance, avoiding US CLOUD Act)
Financial institutions (data sovereignty requirements)
Healthcare systems (HIPAA protection, patient privacy)
Governments (Switzerland, EU nations prioritizing digital independence)
Unions and NGOs (mentioned explicitly in announcement)

Conservative estimate: Enterprise case studies appear Q2 2026 (3-6 months)

Aggressive estimate: Early enterprise adopters deploying Q1 2026 (NOW)

SMB Market: Q3-Q4 2026 (6-9 Months)

Barriers minimal:

Need 4-5 computers (most SMBs already have this)
Need basic IT support (30-minute installation is manageable)
Business case is obvious (data sovereignty + cloud cost savings)

Acceleration factors:

Cloud costs rising (AWS/Azure price increases)
Data breaches increasing (sovereignty argument strengthens daily)
Open-source AI models expanding (more options = more deployment)

My assessment: Fall 2026 for SMB-focused packages and simplified deployment tools

Consumer Market: Late 2026 – Early 2027

Current limitation: “Home users would need more than a single computer” – but hardware evolution is making this feasible rapidly.

Consumer timeline:

Q3-Q4 2026: Tech enthusiasts with home lab setups
Q1-Q2 2027: “Raspberry Pi cluster in a box” consumer products
Q3-Q4 2027: Mass market simplified versions

Historical acceleration patterns:

ChatGPT: Research → API → Consumer = ~12 months
Local LLMs (Llama/Mistral): Expert → Consumer apps = 6-8 months
AI deployment timelines are compressing across the board

The Van Buren Problem (Vote January 14 – 11 Days)

Panattoni’s timeline:

Planning Commission vote: January 14, 2026
If approved, construction: Mid-2026 to 2028
Facility operational: 2028 at earliest

Anyway Systems timeline:

Announced: January 2, 2026
Enterprise deployment: Q2 2026 (6 months)
SMB adoption: Q4 2026 (12 months)
Before Van Buren breaks ground, enterprise alternatives could be standard

The devastating question for Panattoni:

“Your facility won’t be operational until 2028. Anyway Systems is already selling enterprise solutions that will deploy within 6 months. What happens to your business case when distributed computing becomes industry standard before you finish construction?”

This isn’t hypothetical. By the time Van Buren’s facility is operational:

Anyway Systems: 2 years of enterprise deployment
Competitors: Following suit (distributed computing becomes standard practice)
Market shift: Companies choosing sovereignty over centralization
Technology obsolescence: Hyperscale infrastructure built for a model that’s already being replaced

Tax implications:

Van Buren tax exemptions: Through 2050 (25 years)
Anyway enterprise deployment: Q2 2026 (6 months)
Panattoni construction completion: 2028 (2 years)
Technology lifecycle: 5-7 years maximum

Communities are betting 25-year tax subsidies on infrastructure that could be obsolete before it’s built.

Every Michigan Project Faces This Timeline

Project	Construction Timeline	Operational	AnyWAY Enterprise Standard	Risk Window
Oracle/Saline	2026-2027	2027-2028	Q2 2026 (6 months)	1-2 years behind
Van Buren	2026-2028	2028	Q2 2026 (6 months)	2 years behind
Lyon	TBD (2026-2028)	2028+	Q2 2026 (6 months)	2+ years behind
Augusta	If approved 2026	2028+	Q2 2026 (6 months)	2+ years behind

Every project is racing against distributed computing adoption curves. By the time any of these facilities are operational, the technology that makes them unnecessary could be industry standard.

The questions communities should ask:

“Anyway Systems is selling NOW. Enterprise deployment happens Q2 2026. Your facility won’t open until 2028. Explain your business case if distributed computing becomes standard before you’re operational.”
“Technology cycles are 5-7 years. You’re asking for 25-year tax exemptions. Anyway Systems proves alternatives exist. Why should taxpayers subsidize an approach that might be obsolete within your construction timeline?”
“If distributed computing becomes industry standard in 2027 (1 year from now), what happens to promised jobs? To our tax revenue? To your facility?”
“Have you evaluated distributed computing alternatives? If not, why are we approving a multi-billion dollar project without assessing whether it’s necessary?”

This isn’t future speculation. Anyway Systems announced January 2, 2026. Van Buren votes January 14, 2026. Communities have 11 days to demand answers before approving 25-year commitments to potentially obsolete infrastructure.

This Week’s Intelligence

INFRASTRUCTURE DEVELOPMENTS

The Technology That Could Make Hyperscale Data Centers Obsolete

On January 2, researchers at Switzerland’s École Polytechnique Fédérale de Lausanne (EPFL) announced Anyway Systems – software that runs powerful AI models locally instead of in massive remote data centers.

What it does:

Runs ChatGPT-120B on a maximum of 4 computers (not 4 data centers – 4 desktop computers)
Processes AI inference (80-90% of all AI computing) on local networks
Handles “hundreds of billion parameters with just a few GPUs”
Installation takes 30 minutes
Same accuracy as cloud-based systems, slightly slower response time

The researchers’ assessment:

“For years, people have believed that it’s not possible to have large language models and AI tools without huge resources, and that data privacy, sovereignty and sustainability were just victims of this, but this is not the case. Smarter, frugal approaches are possible.“
— Rachid Guerraoui, EPFL DCL Head

How it works: Instead of sending processing to remote servers, Anyway downloads open-source AI models (ChatGPT, Llama, others) to your desktop and distributes processing across a small local network. It self-stabilizes for optimal hardware use and handles failures robustly.

Critical advantages:

Data sovereignty: Private data stays private; companies/governments keep data away from Big Tech
No massive infrastructure: No gigawatt power plants, no millions of gallons of water, no 575 acres of farmland
Scalability: Works with “commodity machines” (normal computers), not specialized data center equipment
Cost: Orders of magnitude cheaper than hyperscale facilities

What this means for Michigan communities:

Saline Township just approved 25-year tax exemptions for Oracle/OpenAI’s $7 billion, 1.4-gigawatt data center. The facility is designed for AI inference – exactly what Anyway Systems proves can run on 4 computers.

Van Buren Township votes January 14 on Panattoni’s 1-gigawatt proposal. Communities should ask: “Swiss researchers proved you can run ChatGPT on 4 computers. Why do you need a gigawatt of power?“

Lyon Township conditionally approved Verrus/Alphabet’s 1.8 million square foot facility in September. Residents discovering it now should demand answers: “If distributed computing works, why wasn’t this considered?”

The Questions Communities Should Ask Developers

On Technology Necessity:

“EPFL researchers run ChatGPT-120B on 4 computers using Anyway Systems. Why does your project require 1.4 gigawatts of power and 575 acres?”
“What percentage of your computing is AI inference (which Anyway Systems handles on local networks)?”
“Have you evaluated distributed computing alternatives? If not, why are taxpayers subsidizing the most expensive approach?”
“If Swiss researchers can achieve the same results with frugal approaches, what prevents you from using them?”

On Financial Risk:

“Technology lifecycles are 5-7 years. Tax exemptions run through 2050 (25 years). What happens if distributed computing becomes standard and your facility is obsolete by 2030?”
“Will communities still receive property tax revenue if you close or downgrade the facility before 2050?”
“Google’s Oregon data center closed after 15 years when tax exemptions expired. Do you have similar plans?”

On Long-Term Value:

“If we’re subsidizing infrastructure that might become obsolete, shouldn’t communities own equity rather than just granting tax breaks?”
“Swiss solutions keep data sovereign and private. Why are we subsidizing centralized infrastructure controlled by Big Tech?”
“Four computers require minimal power and water. Your facility requires a gigawatt and millions of gallons daily. Is this the best use of Great Lakes resources?”

The Timing Is Critical

Van Buren Township: January 14, 2026 vote (11 days from now)
Lyon Township: Conditionally approved but sound study and energy requirements still pending
Augusta Township: Ballot referendum expected 2026
Saline Township: Construction beginning but legal challenges pending

Communities still have leverage. Anyway Systems proves the “we need massive infrastructure” argument isn’t absolute truth – it’s a business model choice that benefits developers, not necessarily communities.

Sources: New Atlas, EPFL, Anyway Systems

Oracle Breaks Ground Despite Financial Instability

Despite Blue Owl Capital’s withdrawal from Saline financing (reported December 18), Oracle began site preparation in early January. Gravel trucks, DTE trucks blocking lanes, and constant backup alarm noise confirm construction is underway.

The financial concern pattern:

December 11: Oracle posts negative $10B free cash flow, stock drops 11%
December 12: Bloomberg reports 2027→2028 delays across multiple OpenAI data centers
December 18: Blue Owl Capital exits Saline financing same day as MPSC approval
December 19: Financial Times confirms equity partner change (replacement unnamed)
January 1-3: Oracle breaks ground without publicly naming new equity partner

What’s missing: Related Digital claims a “new equity partner” replaced Blue Owl but refuses to identify them. For a $7 billion project, this secrecy raises questions:

Who has the capital to replace Blue Owl on short notice?
Why won’t they announce the partner if financing is solid?
Is Oracle funding more of the project directly (increasing their financial risk)?
What happens if Oracle’s credit rating gets downgraded to junk (as Moody’s and S&P have suggested)?

Community protection gap: Michigan Public Service Commission Chair Dan Scripps emphasized contracts include “upfront collateral” and “don’t assume the customer will stay financially solvent.” But if the unnamed equity partner fails to materialize or Oracle restructures mid-construction:

Tax breaks are already committed (no clawback provisions exist)
DTE grid upgrades could be stranded costs
Promised jobs never materialize with no recourse
Communities face specialized infrastructure designed for one use

Legal challenges continue: Saline resident filed legal intervention alleging township officials violated Open Meetings Act during settlement approval. Township Supervisor Jim Marion, the only board member who voted against the settlement: “I was trying to stick up for the township residents. If we would have put it on the ballot, it would have been overwhelmingly against it.”

Sources: MLive, Planet Detroit, Bloomberg, Financial Times, Wall Street Journal

Bipartisan Push to Repeal ALL Data Center Tax Incentives

On December 16, Representative Dylan Wegela (D-Garden City) and 11 bipartisan co-sponsors introduced legislation to repeal Michigan’s data center tax breaks entirely.

Current tax structure:

6% sales/use tax exemption on equipment through 2050 (greenfield sites)
Through 2065 for brownfield sites (25-40 years total)
PA 198 (50% additional property tax abatement) available on request
No job creation requirements
No clawback provisions if facilities close early
No performance guarantees

Critical discovery: These incentives have ZERO accountability mechanisms. If a data center:

Closes after 10 years (tax exemptions run 25 years)
Never creates promised jobs
Gets sold to another company
Enters bankruptcy

Communities have no recourse. Tax breaks are permanent, jobs are promises, and there’s no clawback language.

The numbers:

$13 million in tax breaks claimed so far (as of December 2025)
$42.5 million estimated through 2062 (Senate Fiscal Agency analysis)
“Expected to grow exponentially” with ~20 projects in pipeline
Money affects School Aid Fund (K-12 schools), general fund, and local municipalities

The 25-year gamble:

Saline example:

Investment: $7 billion
Tax exemptions: Through 2050 (25 years)
Technology lifecycle: 5-7 years for servers, 10-15 years max for facilities
Anyway Systems: Proves same AI runs on 4 computers (announced January 2)

Critical question: What happens in 2030 when distributed computing becomes standard, but tax exemptions run through 2050?

Likely scenarios:

Facility closes/downgrades → Community gets specialized infrastructure unusable for other purposes + zero tax revenue for remaining 20 years
Ownership changes → New owner inherits tax breaks, community loses negotiating leverage
Oracle restructures → Facility continues but under different terms, jobs disappear
Technology shifts → Hyperscale becomes obsolete, facility repurposed (if possible)

None of these scenarios include clawback provisions. Communities bet 25 years of tax revenue on 5-7 year technology.

Political momentum:

Michigan House Speaker Matt Hall (R): Expressed “openness” to repealing tax breaks
19 state attorneys general: Warned AI companies about consumer protection violations (December)
Bipartisan opposition: Republicans and Democrats both sponsoring repeal
Community organizing: “Michiganders Against Data Centers” holding rallies, 800+ attended December MPSC hearing

What’s at stake: If repeal passes, existing projects keep their exemptions (grandfather clause expected) but new projects lose subsidies. This could:

Stop the pipeline of 15+ proposed Michigan projects
Force better negotiations (developers need incentives less than communities need leverage)
Redirect $42.5M+ toward schools, infrastructure, or alternatives

Watch for: House floor vote timing (likely Q1 2026). If passed before Van Buren vote (January 14), that project loses tax subsidy eligibility.

Sources: Bridge Michigan, Inside Climate News, Michigan Advance, Planet Detroit

Michigan Strategic Fund Creates Tax Loophole in Clean Energy Requirements

While legislators push to repeal tax breaks, the Michigan Strategic Fund (MSF) quietly undermined existing protections. In November 2025, MSF issued guidelines stating data centers qualify for tax exemptions if they “contract with a utility provider subject to Michigan’s clean energy regulatory framework.”

The problem: No Michigan utility has 90% clean energy now or will by 2031 (6-year deadline). Simply contracting with DTE or Consumers Energy doesn’t mean the data center uses 90% clean energy – it means they use whatever mix the utility provides (currently ~40% fossil fuels).

The law says: Data centers must prove they procure – or will procure within 6 years – 90% of power from clean/renewable sources.

MSF’s interpretation: Contracting with any Michigan utility automatically satisfies this requirement.

The loophole: This renders the 90% clean energy requirement meaningless. Every data center can claim compliance without actually using clean energy.

Who called it out:

Michigan Environmental Council
NRDC (Natural Resources Defense Council)
Sierra Club
Citizens Utility Board of Michigan
Troposphere Legal

December 2025 letter to MSF: Organizations demanded revisions, stating MSF guidelines “do not adequately enforce clean energy requirements enshrined in Michigan state law.”

Why this matters: Data centers are already incentivized to use fossil fuels (cheaper, more reliable, faster to deploy). The 90% clean energy requirement was supposed to force them toward renewables. MSF’s loophole means they can build gas plants instead, locking in fossil fuel infrastructure for decades.

Community leverage point: Demand developers prove actual clean energy procurement, not just utility contracts. Ask: “How will you achieve 90% clean energy when DTE is only 40% clean?”

Sources: NRDC, Michigan Environmental Council, Sierra Club, Citizens Utility Board of Michigan

POLICY DEVELOPMENTS

Attorney General Dana Nessel Challenges Consumers Energy Data Center Rules

Michigan Attorney General Dana Nessel is challenging Michigan Public Service Commission rules governing data centers powered by Consumers Energy, calling them inadequate for protecting ratepayers.

This follows her earlier intervention in DTE/Oracle/OpenAI contracts, which the MPSC approved via fast-track ex parte process over her objections.

Nessel’s concern: Michigan’s data center legislation promises ratepayer protection but provides vague language with limited enforcement.

The regulatory gap: Douglas Jester (consultant, 5 Lakes Energy) warned in October: “The existing framework for allocating costs could leave residential ratepayers and other customers paying for a portion of the electric generation needed solely to power large data centers.”

The compounding problem: Demand from data centers without sufficient generation drives wholesale power cost increases that affect all customers. Michigan’s utilities are already stretched:

DTE total capacity: ~18.6 GW
Known data center requests: 2.4+ GW (Saline 1.4GW + Van Buren 1GW alone)
That’s 13% capacity increase from just two projects

Attorney General’s leverage: Nessel’s challenges create legal uncertainty that could delay approvals or force better ratepayer protections. Watch for MPSC response in Q1 2026.

Sources: Planet Detroit, Michigan Advance

Great Lakes Water Levels: Still Below Long-Term Averages

December 2025 water level update (latest available data):

All lakes: 2-5 inches below last month
All except Erie: 4-13 inches below long-term December averages
All lakes: 14-28 inches above record lows (but trending down)

Year-over-year comparison (December 2025 vs. December 2024):

Lake Michigan-Huron: 8 inches below
Lake St. Clair: 5 inches below
Lake Erie: 7 inches below
Lake Superior: Near last year’s level
Lake Ontario: 1 inch higher

Decade comparison (International Joint Commission February 2025 analysis – most recent comprehensive report):

Lake Superior: Lowest since 2013
Lake Michigan-Huron: Lowest since 2014
Lake Ontario: Lowest since 2003
Lake Erie: Lowest since 2014 (but ~0 inches above long-term average)

What caused accelerated decline: Drought conditions in certain Great Lakes regions accelerated natural seasonal water level decline throughout fall 2024 and into winter 2025.

The data center timing crisis:

Michigan is approving facilities requiring:

Hyperscale (like Saline): Up to 5 million gallons daily = 1.8 billion gallons annually
Medium facilities: 110 million gallons annually
Multiple projects: 15+ proposals across Michigan

All while Great Lakes hit decade lows.

Developer claims vs. reality:

Saline (Oracle/OpenAI): “Closed-loop cooling = comparable to office building”
Lyon (Verrus/Alphabet): “15,000 gallons/day vs. typical 5.25M gallons/day for this size”

Community verification needed: Demand third-party engineering verification of water consumption claims. Require:

Real-time water consumption monitoring
Curtailment requirements if Great Lakes levels fall further
Financial penalties for exceeding permitted amounts
Posted bonds for emergency drought response
Comparison to AnyWAY Systems alternative (4 computers = minimal water use)

Sources: International Joint Commission, U.S. Army Corps of Engineers, NOAA GLERL, Planet Detroit

TECH EMPLOYMENT SNAPSHOT

2025 Tech Layoffs: Final Count

Multiple tracking sources confirm final 2025 numbers:

Layoffs.fyi: 122,549 tech employees at 257 companies
TrueUp.io: 209,838 people at 716 companies
Crunchbase: 126,352 U.S. tech workers in mass layoffs

2025 vs. 2024:

2025: 209,838 workers (down 12% from 2024)
2024: 239,101 workers
Daily average 2025: 583 layoffs/day
Daily average 2024: 655 layoffs/day

The pattern: Layoffs decreased but remain historically elevated. More importantly, companies are deploying AI capabilities (Meta’s $2B Manus acquisition, OpenAI’s voice AI, enterprise agent platforms) while maintaining headcount – the reduction comes later through attrition and “reorganizations.”

Major 2025 cuts:

Amazon: 14,000+ announced
Intel: Reducing to 75,000 employees (15-20% of foundry division)
Microsoft: 9,000 employees (~4% workforce)
Verizon: 13,000+ (largest-ever campaign)
Salesforce: ~5,000 roles

The disconnect: Tech companies laid off 209,838 workers in 2025 while simultaneously:

Announcing $121B in data center bonds
Promising thousands of construction jobs
Claiming AI creates employment opportunities
Building infrastructure that automates human work

For deeper employment intelligence: See this week’s Under the Radar newsletter covering Meta’s $2B Manus acquisition (AI agents that automate administrative work), OpenAI’s voice AI timeline, and immediate impacts on administrative and customer service roles.

Sources: Layoffs.fyi, TrueUp.io, Crunchbase, TechCrunch, Challenger Gray & Christmas

MICHIGAN INFRASTRUCTURE TRACKER

16+ Active Data Center Projects (Updated January 2026)

Project	Location	Power	Investment	Status	Tax Subsidy	Cost/Job	Obsolescence Risk
Oracle/OpenAI Stargate	Saline	1.4 GW	$7B	Breaking ground	~$420M (2050)	$215K-$933K	HIGH – Anyway proves alternative
Panattoni	Van Buren	1 GW	$4-7B est	Vote Jan 14	Would qualify 6% + PA 198	$6-10M est	HIGH – AI inference focus
Verrus/Alphabet	Lyon	Unknown	$4B est	Approved Sept ’25	Would qualify 6%	$40-80M	HIGH – 1.8M sq ft hyperscale
Microsoft/OpenAI	Southfield	100 MW	Unknown	Approved Dec ’25	Would qualify 6%	$3-5M est	MEDIUM
U-M/Los Alamos	Ypsilanti	1.2 GW	$1.2B	Grant controversy	N/A (research)	Unknown	MEDIUM – research focus
Thor Equities	Augusta	100 MW	$1B	Ballot 2026	Would qualify	TBD	MEDIUM
Sansone Group	York	1 GW est	Unknown	Residents seeking moratorium	Would qualify	TBD	HIGH if hyperscale
Microsoft	Gaines	Unknown	Unknown	Rezoning requested	Would qualify	TBD	TBD
Microsoft	Dorr	Unknown	Unknown	272 acres acquired 2024	Would qualify	TBD	TBD
Franklin Partners	Lowell	Unknown	Unknown	Public hearing Jan 12	Would qualify	TBD	TBD
Deep Green	Lansing	24 MW	$120M	Downtown project	Different model	TBD	LOW – waste heat model
Cloverleaf	Frenchtown	Unknown	Unknown	Early outreach	Would qualify	TBD	TBD
Cloverleaf	Dundee	Unknown	Unknown	Moratorium enacted	Would qualify	TBD	TBD

NEW COLUMN ADDED: “Obsolescence Risk” based on Anyway Systems announcement. Projects focused on AI inference face HIGH risk of technology obsolescence before tax subsidies expire.

Total known capacity requests: 5+ gigawatts across active proposals
DTE total capacity: ~18.6 GW
Consumers Energy capacity: ~7.6 GW
Combined data center pipeline: ~22 GW being discussed

AnyWAY Systems impact: Every project claiming “we need massive infrastructure for AI” now faces the question: “Swiss researchers proved otherwise – why can’t you use distributed computing?”

WHAT TO WATCH

Next 14 Days:

January 14, 2026 – Van Buren Township Planning Commission Vote (Critical)

1-gigawatt Panattoni proposal
Community opposition strong
5:30 PM, Board of Trustees Room, Township Hall, 46425 Tyler Road
Key questions to ask:
- “Anyway Systems runs ChatGPT on 4 computers. Why do you need 1 gigawatt?”
- “What percentage of your computing is AI inference (which distributed computing handles)?”
- “Have you evaluated alternatives? If not, why are taxpayers subsidizing the most expensive option?”

January 12, 2026 – Lowell Township Public Hearing

Franklin Partners 235-acre proposal
Watch for “permitted by right” arguments
Community organizing opportunity

Q1 2026 Developments:

Tax repeal vote: House floor vote on bipartisan repeal legislation (timing unclear but momentum building)
Van Buren aftermath: If approved January 14, watch community organizing response
Lyon informational meeting: Late January (4 months after September approval)
Oracle Q3 earnings: Late January/early February – financial stability indicators
Credit rating actions: Moody’s and S&P decisions on Oracle debt rating

Longer-Term (Q1 2026):

The AnyWAY Systems Question:

Will developers address distributed computing alternatives?
Will communities demand evaluation of cheaper, less resource-intensive options?
Will Michigan Strategic Fund revise guidance based on new technology?
Will tax repeal gain momentum as alternative technologies emerge?

Grid Capacity Reality Check:

Can DTE/Consumers deliver 2.4+ GW new capacity without rate increases?
Will MPSC require utilities to build new gas plants (violating clean energy goals)?
What happens when distributed computing reduces demand projections?

Water Permits During Drought:

Will EGLE approve millions of gallons daily when Great Lakes at decade lows?
Can developers prove “closed-loop” claims with third-party verification?
Will drought conditions worsen (strengthening community leverage)?

Employment Verification:

When Saline construction starts, do promised jobs materialize?
What percentage are local vs. traveling crews?
How many permanent vs. construction (temporary)?

Community Organizing Effectiveness:

Can townships use Anyway Systems intelligence to stop or negotiate better terms?
Will moratorium strategy (Howell, Dundee) spread?
Can communities demand equity ownership instead of just tax breaks?

DATA YOU CAN USE

The 25-Year Tax Subsidy vs. 5-Year Technology Lifecycle

Scenario	Tax Exemption Period	Technology Lifecycle	Risk to Community
Saline (Oracle)	Through 2050 (25 years)	5-7 years servers, 10-15 max facility	If closed by 2035, community loses 15 years tax revenue
Van Buren	Through 2050 (25 years)	Same	If closed by 2035, community loses 15 years tax revenue
Lyon (brownfield)	Through 2065 (40 years)	Same	If closed by 2035, community loses 30 years tax revenue

Historical precedent: Google’s Oregon data center – closed/downgraded after tax exemptions expired (15 years). No evidence they intended to operate beyond subsidy period.

Anyway Systems impact: If distributed computing becomes standard by 2030, hyperscale facilities may close/repurpose while tax exemptions continue.

Community protection: ZERO clawback provisions exist. If facilities close early, communities have no recourse.

Distributed Computing vs. Hyperscale Infrastructure

Approach	Power Required	Water Use	Land Use	Investment	Technology Risk
Anyway Systems (Distributed)	Minimal (4 computers)	Negligible	Existing office space	Software + commodity hardware	LOW – uses existing infrastructure
Saline (Hyperscale)	1.4 GW	Up to 5M gal/day claimed	575 acres farmland	$7 billion	HIGH – single-purpose infrastructure
Van Buren (Hyperscale)	1 GW	Millions gal/day	282 acres	$4-7B estimated	HIGH – single-purpose infrastructure

The question communities should ask: “Why subsidize the expensive option when alternatives exist?”

Community Negotiation Leverage Points (Updated for Anyway Systems)

Technology Alternatives:

“EPFL proved ChatGPT runs on 4 computers. Have you evaluated distributed computing?”
“What percentage of your operations could use Anyway-style systems?”
“If technology shifts to distributed computing in 5 years, what happens to your facility?”
“Why should taxpayers subsidize hyperscale when frugal alternatives exist?”

Financial Stability (Oracle specifically):

“Who replaced Blue Owl Capital as equity partner? Why won’t you name them?”
“What happens if Oracle’s credit rating drops to junk status?”
“Bloomberg reported delays across your portfolio. Why should we believe Saline timeline?”
“Your stock is down 15% from recent highs. Is this a good financial bet for our community?”

Long-Term Value:

“Tax exemptions run 25 years. Technology cycles are 5-7 years. What’s our protection?”
“If you close early, do we get tax revenue back? (Answer: No clawbacks exist)”
“Google closed Oregon data center after exemptions expired. Will you do the same?”
“Shouldn’t communities own equity instead of just granting tax breaks?”

Environmental Protection:

“Great Lakes are at decade lows. How will you curtail water use during drought?”
“Closed-loop cooling claims need third-party verification. Will you provide it?”
“Anyway Systems uses negligible water. Why do you need millions of gallons daily?”
“What’s your plan if EGLE denies water permits due to environmental conditions?”

Job Creation Reality:

“Construction jobs are temporary (18-24 months). What are permanent job numbers?”
“Local hiring percentage with financial penalties for non-compliance?”
“Independent verification of job creation (not developer self-reporting)?”
“If AI can run on 4 computers, how many operational jobs do you actually need?”

INTELLIGENCE ANALYSIS

The AnyWAY Systems Paradigm Shift

This week’s most significant development isn’t a Michigan project update or policy change. It’s the proof that the entire hyperscale data center model isn’t necessary.

For 3 years, communities have heard:

“AI requires massive infrastructure”
“We need gigawatts of power”
“Data centers are essential for AI progress”
“There’s no alternative to hyperscale”

Swiss researchers just proved otherwise. ChatGPT-120B runs on 4 computers. Not 4 data centers. Four normal computers on a local network.

The implications cascade:

1. Technology Obsolescence Risk Is Real

Communities are betting 25-year tax exemptions on 5-7 year technology. If distributed computing becomes standard:

Scenario A: Hyperscale facilities become stranded assets, communities get specialized infrastructure with no alternative use
Scenario B: Facilities repurpose (if possible), but not for promised AI work – jobs disappear
Scenario C: Ownership changes, new owners inherit tax breaks, community loses leverage
Scenario D: Facilities close outright, communities get nothing for remaining exemption years

None of these scenarios include clawback provisions. The risk is entirely on communities.

2. The “Necessity” Argument Collapses

Developers claim they “need” hyperscale infrastructure. Anyway Systems proves this is a business model choice, not technical necessity.

Business model comparison:

Hyperscale (Current):

Centralized control (Big Tech owns infrastructure)
Data leaves user control (privacy concerns)
Massive capital investment ($7B+ facilities)
Ongoing operational costs (power, cooling, maintenance)
Vendor lock-in (users depend on specific platforms)

Distributed (Anyway):

Local control (companies/governments own their infrastructure)
Data stays sovereign (privacy protected)
Minimal capital investment (commodity hardware)
Lower operational costs (no massive cooling, power)
Platform flexibility (open-source models, interchangeable)

Why choose hyperscale? Not because it’s technically necessary. Because it’s more profitable for developers and creates vendor lock-in for customers.

Communities subsidizing the profitable choice, not the necessary one.

3. The Timing Creates Leverage

Van Buren votes January 14 – 11 days from now. Anyway Systems was announced January 2. Van Buren Planning Commission has cutting-edge intelligence that undermines Panattoni’s entire argument.

Questions Van Buren should ask:

“We learned this week that Swiss researchers run ChatGPT on 4 computers. Your proposal requires 1 gigawatt of power. Explain the discrepancy.”
“Have you evaluated distributed computing alternatives? If not, why are taxpayers subsidizing an approach that might become obsolete?”
“If Anyway-style systems become standard in 3 years, what happens to your facility? To promised jobs? To our tax revenue?”
“Your business model depends on centralized infrastructure. Distributed computing threatens that model. Why should we bet our community’s future on your obsolete approach?”

Lyon, Augusta, York, and other pending projects: Same questions apply. Demand evaluation of distributed alternatives before approving subsidies.

4. The Clean Energy Loophole Makes More Sense Now

Michigan Strategic Fund created a loophole allowing data centers to claim “90% clean energy” compliance just by contracting with utilities – even though no Michigan utility is 90% clean.

Why would MSF do this? Because data centers want to build gas plants (cheaper, more reliable, faster). The 90% requirement threatened their business model.

But if distributed computing becomes standard: Those gas plants become stranded assets. Michigan locks in fossil fuel infrastructure for decades… for data centers that might not need it in 5 years.

The environmental groups are right to oppose this. Not just on climate grounds, but on economic grounds. We’re subsidizing infrastructure that may become obsolete before it’s paid off.

The Financial Instability Pattern

Oracle’s Saline project continues raising red flags:

Pattern across Oracle’s portfolio:

Negative $10B free cash flow (Q2 2025)
$248B in lease commitments for unbuilt facilities
Blue Owl Capital exit from Michigan financing
Bloomberg reporting 2027→2028 delays
Stock down 15% from highs
Credit rating agencies moving toward junk status
Construction starting despite all of this

Why start construction with unstable financing?

Possible explanation: Once construction begins, communities have sunk costs (infrastructure damage, legal settlements, MPSC approvals). Oracle gains leverage. “We’ve already started – stopping now costs you more than letting us continue.”

The unnamed equity partner: Related Digital refuses to identify Blue Owl’s replacement. For a $7B project, this is unusual. Possibilities:

Oracle itself: Funding more directly (increasing their financial risk)
Weaker partner: Less stable than Blue Owl, don’t want scrutiny
Terms unclear: Still negotiating, no firm commitment
No partner yet: Breaking ground to create pressure/leverage

None of these scenarios suggest strong financial stability.

Community exposure: Saline committed tax breaks, settled lawsuit, approved power contracts. If Oracle’s financing collapses:

DTE already investing in grid upgrades (ratepayers paying even if project fails)
Township already damaged roads with heavy trucks (repair costs fall on township)
Legal settlement already paid $14M in community concessions (no clawback if project fails)
Tax exemptions already committed through 2050 (no revenue recovery if facility closes early)

This is exactly why communities need protection mechanisms: Performance bonds, escrow accounts, clawback provisions, financial stability verification. None of these exist in Michigan’s data center legislation.

The Tax Repeal Momentum

Bipartisan opposition to data center tax breaks is unusual. Republicans and Democrats rarely agree on economic development incentives. What changed?

The coalition:

Progressive Democrats: Environmental concerns, climate impact, corporate welfare
Conservative Republicans: Fiscal responsibility, property rights, rural character preservation
Libertarians: Government picking winners, market distortion
Community activists: Water use, farmland loss, noise, traffic

This crosses traditional political boundaries. It’s not “left vs. right” – it’s “communities vs. developers.”

Michigan House Speaker Matt Hall (R) expressing “openness” to repeal is significant. Hall voted against the original tax breaks. His willingness to consider repeal suggests:

Political calculation: Constituent opposition is strong enough to overcome developer lobbying
Fiscal concerns: $42.5M+ in lost revenue affecting schools is politically risky
Technology uncertainty: Anyway Systems and similar developments make long-term subsidies questionable

If repeal passes:

Existing projects keep exemptions (grandfather clause expected – can’t retroactively change approved deals)
New projects lose subsidies (Van Buren, Lyon, Augusta, others)
Negotiating leverage shifts dramatically (developers need communities more than communities need developers)

Communities should:

Delay approvals until legislative uncertainty resolves (use moratoriums)
Demand better terms given subsidy uncertainty (equity ownership, performance bonds, job guarantees)
Leverage Anyway Systems (“Why subsidize expensive approach when alternatives exist?”)

Timeline: If House votes before January 14, Van Buren loses eligibility. If House votes after January 14 but project isn’t approved yet, Van Buren could still lose eligibility. Timing matters.

COMMUNITY NEGOTIATION TOOLKIT (Updated)

The Anyway Systems Argument

Core principle: “Prove necessity before granting subsidies”

Questions for developers:

Technology Evaluation:

“Swiss researchers at EPFL announced January 2 that AnyWAY Systems runs ChatGPT-120B on 4 computers. Your proposal requires [X gigawatts]. Explain why distributed computing won’t work for your use case.”
“What percentage of your computing workload is AI inference (the 80-90% of AI computing that Anyway Systems handles on local networks)?”
“Have you conducted an internal evaluation of distributed computing alternatives? If yes, share results. If no, explain why taxpayers should subsidize an approach without evaluating alternatives.”
“Anyway Systems provides data sovereignty (data stays local), reduced environmental impact (minimal power/water), and lower cost. What advantages does your hyperscale model provide communities that offset these benefits?”
“If distributed computing becomes industry standard in 3-5 years, what happens to your facility? To promised jobs? To our tax revenue through 2050?”

Financial Stability (Especially for Oracle/Related Digital Projects):

“Blue Owl Capital withdrew from Saline financing December 18. Who replaced them as equity partner? Why won’t you publicly name them?”
“Oracle posted negative $10 billion free cash flow in Q2 2025 and delayed multiple projects. Provide evidence of financial capacity to complete this project.”
“Moody’s and S&P are considering downgrading Oracle bonds to junk status. What’s your contingency if Oracle’s credit rating drops mid-construction?”
“Bloomberg reported you’re delaying 2027 projects to 2028 due to ‘labor and material shortages.’ Why should we believe our timeline is different?”
“Will you post a performance bond equal to total tax subsidies, refundable if you complete the project but forfeited if you don’t?”

Long-Term Community Protection:

“Tax exemptions run through 2050 (25 years). Data center technology cycles are 5-7 years. Include clawback provisions: if you close before 2050, you refund proportional tax subsidies.”
“Google’s Oregon data center closed after tax exemptions expired. Include contractual commitment to operate through 2050 or refund subsidies.”
“Instead of tax exemptions, offer communities equity ownership. We take financial risk – we should get financial upside.”
“Include job creation guarantees with financial penalties: if you promise 450 jobs but only create 200, you pay [$X] penalty per missing job.”
“Include technology obsolescence protection: if distributed computing makes your facility obsolete, you fund community transition costs.”

Environmental Verification:

“Great Lakes are at decade lows (Lake Superior lowest since 2013). Include mandatory water curtailment during drought conditions.”
“Prove ‘closed-loop cooling’ claims with third-party engineering verification, not developer promises.”
“Anyway Systems uses negligible water. Justify why your approach requires millions of gallons daily when alternatives exist.”
“Include real-time water monitoring with public reporting and automatic fines for exceeding permitted amounts.”
“Post environmental remediation bond: if you contaminate water or soil, bond covers cleanup costs.”

Job Creation Reality:

“Distinguish construction jobs (temporary, 18-24 months) from operational jobs (permanent). We want permanent job numbers, not inflated construction numbers.”
“What percentage of construction jobs go to local residents vs. traveling specialty contractors? Include local hiring requirements with penalties.”
“If Anyway Systems proves AI runs on 4 computers, how many operational jobs do you actually need? Show us the math.”
“Include independent job creation verification (state agency audit, not developer self-reporting).”
“Define ‘jobs created’ clearly: full-time equivalent, salary ranges, skill requirements. No counting 1 part-time contractor as 1 ‘job.'”

The Moratorium Strategy

Proven effective:

Howell Township: 6-month moratorium → Meta withdrew proposal
Dundee Township: Moratorium enacted → developer paused
Pavilion Township: Community backlash → developer withdrew

How to implement:

Township board votes moratorium on data center development (6-12 months)
Use time to:
- Study Anyway Systems and distributed alternatives
- Develop zoning amendments requiring performance standards
- Wait for tax repeal legislation to resolve
- Organize community education and input
- Consult with townships that rejected projects (Howell, Dundee)
Strengthen negotiating position: Moratorium shows developers community has leverage, not desperation

Legal risk: Developers threaten lawsuits (like Saline). But:

Howell faced no lawsuit (Meta just left)
Dundee faced no lawsuit (developer paused)
Saline settled because township was vulnerable (rushed rezoning, weak legal position)

Strong moratoriums with clear rationale (public safety, infrastructure capacity, environmental review) are legally defensible.

The “We Need Better Information” Argument

Current situation: Developers provide:

Promotional materials (glossy promises)
Estimated job numbers (often inflated)
Vague environmental claims (unverified)
Self-serving financial projections (no independent validation)

What communities should demand:

Third-party impact analysis (not developer-funded consultants)
Comparable project data (actual performance of existing data centers, not projections)
Technology evaluation (distributed computing feasibility assessment)
Financial stability verification (credit ratings, equity partner identification, performance bonds)
Environmental engineering (water use verification, noise studies, traffic impact)

The “we need time to study” argument is legally bulletproof. Courts don’t force communities to approve projects without adequate information.

Timeline: Van Buren could vote “no” on January 14 with justification: “We need time to evaluate distributed computing alternatives announced January 2. Six-month moratorium for technology review.”

This isn’t obstruction. This is due diligence.

RESOURCES

Intelligence Tracking:

PivotIntel App: pivotintel.org/app – Real-time project monitoring
Great Lakes Dashboard: glerl.noaa.gov/data/wlevels/dashboard
Army Corps Water Levels: lrd.usace.army.mil/Water-Information
IJC Updates: ijc.org/en/great-lakes-water-levels-boards

This Week’s Key Sources:

Anyway Systems: anyway.dev – Technology details
EPFL Announcement: actu.epfl.ch
New Atlas Coverage: newatlas.com

Township Meetings:

Van Buren: January 14, 2026, 5:30 PM, Township Hall, 46425 Tyler Road
Lowell: January 12, 2026 (public hearing on Franklin Partners 235-acre proposal)
Lyon: Late January 2026 informational meeting (date TBD)

Community Organizing:

Study moratorium successes (Howell, Dundee, Pavilion)
Use Anyway Systems intelligence as negotiating leverage
Demand technology alternatives evaluation before approval
Share information across townships facing similar proposals
Document developer promises vs. reality for future negotiations

Cross-Reference:

For worker impact and career intelligence: Under the Radar – January 3, 2026

Meta’s $2B Manus acquisition (administrative work automation)
OpenAI’s voice AI timeline (customer service automation)
Forward Deployed Engineers opportunity (implementing AI systems)
Healthcare Patient Care Coordinators (automation-resistant)
Systems thinking skills framework (Week 4 of foundation skills series)

Analysis & Background:

Under the Radar: theopenrecord.org/under-the-radar – Career intelligence
Beyond Hyperscale Alternatives: theopenrecord.org/2025/12/15/beyond-hyperscale
AI Deployment vs. Bubble: theopenrecord.org – Multi-part series

METHODOLOGY

Intelligence gathered from:

Technology Sources: New Atlas, EPFL, AnyWAY Systems, TechCrunch
Michigan Regional: Planet Detroit, MLive, Bridge Michigan, Detroit News, Michigan Advance, WXYZ, FOX 2 Detroit
Financial: Bloomberg, Financial Times, Wall Street Journal
Government: Michigan Public Service Commission filings, EGLE permits, International Joint Commission, NOAA GLERL, U.S. Army Corps of Engineers
Legal: Attorney General statements, court filings, township records
Employment: Layoffs.fyi, Crunchbase, TechCrunch, Challenger Gray & Christmas

All major claims backed by primary sources. Water level data from IJC, NOAA GLERL, and U.S. Army Corps of Engineers official monitoring networks. Financial analysis from publicly available corporate filings and credit rating agency reports. Technology assessment from peer-reviewed research institution (EPFL).

Estimates clearly identified. Cost-per-job calculations show methodology. Confidence scoring available in PivotIntel app for individual project data.

Questions? Intelligence tips? Community organizing stories?
Email: angela@theopenrecord.org

About PivotIntel: Infrastructure intelligence for communities navigating AI development. Focus: data centers, power grid, water resources, employment reality, financial stability, community leverage.

Published by The Open Record L3C | theopenrecord.org

Next Edition: Sunday, January 12, 2026

All sources archived via Wayback Machine for permanent verification.