Pablo Zavala · AI Safety Evaluation · Research Engineering

Gigabit City: Evidence, Equity, and the BEAD Playbook

A four-page evidence memo distills the empirical record on U.S. gigabit fiber, grading every claim as causal or descriptive: Chattanooga's $2.69 billion municipal payoff, a roughly 20% incumbent speed response to Google Fiber entry, 2% and nearly 9% housing premiums, speed thresholds for business births, recurring adoption gaps, and a deployment playbook BEAD should enforce.

July 10, 2025 · 6 min

Google pitched the gigabit as a "market catalyst," a spark to prompt faster, cheaper broadband. Fifteen years on, I wrote a four-page evidence memo at Carnegie Mellon's Block Center asking what reaction the empirical record documents. The memo, hosted here as Gigabit City Analysis (PDF), synthesizes published evaluations of the pioneer markets, Kansas City, Chattanooga, Austin, and the Research Triangle, and extracts a deployment playbook from Provo, West Des Moines, and Louisville. This essay walks through what survives scrutiny.

A boundary belongs up front: the memo synthesizes other researchers' published studies and reports zero original causal estimates of its own. Every figure below therefore carries the methods and limits of its cited study and stays checkable against the sources at the end. One rule governs the memo and threads this essay: treat an effect as causal only when rigorous empirical work documents it, and label everything else descriptive.

What Rigorous Studies Support

Run the record through that rule and Chattanooga anchors what survives. Years before Google Fiber existed, its municipal utility, EPB, built citywide fiber and treated broadband as an essential public service. A ten-year evaluation estimated $2.69 billion in economic benefit and 9,516 jobs created or saved, with economic development driving 52% of the value and smart-grid efficiencies another 28% (Lobo 2020); residential take-up reached 58% by 2020. Yet even the anchor carries a flag: the dollar and job figures derive from an input-output model, gross estimated flows rather than net employment change.

Competition supplies the cleanest entry-effect result: where Google Fiber launched, incumbents raised average advertised download speeds by approximately 20%, about 47 Mbps (Solis 2023). Austin shows the dynamic at street level, average speeds climbing from 28 Mbps in 2012 to 144 Mbps by 2018 as AT&T and Grande Communications accelerated their own gigabit buildouts (GovTech 2020). Here the labeling rule earns its keep: the memo treats the Solis estimate as rigorous and the Austin trajectory as descriptive, a separation its summary table enforces cell by cell.

Housing evidence arrived in 2024: fiber availability added a home-price premium of 2% in Minnesota markets and nearly 9% in Texas markets (Whitacre 2024). City-specific hedonic studies for the pioneer metros remain missing, so the memo quotes the two-state result and leaves the metro cells empty; borrowed precision would flatter the record.

The sharpest policy finding turns the rule on the gigabit brand itself. An eight-state event study found 100+ Mbps and 250 Mbps introductions increased business births, while evidence for an additional gigabit-level boost stayed inconclusive across 2015 to 2020 (Biedny et al. 2024). Speed fuels entrepreneurship up to a threshold; past it, the marquee number buys marketing value more than measured firm creation.

Equity Gaps Repeat Across Models

Strong effects settle half the policy question; adoption settles the rest, and there the record repeats. Kansas City pioneered demand aggregation: neighborhoods, branded as fiberhoods, qualified for construction by hitting pre-registration targets, and 180 of 202 qualified, nearly 90%, atop a first phase estimated at $84 million to pass 149,000 homes (Google FCC 2015; CostQuest 2020). Yet the same design entrenched a divide: take-up in some lower-income neighborhoods eventually reached 30% while higher-income areas hit 75% (State of Connecticut 2022).

Austin reproduced the gap through a different mechanism: the Community Connections program promised free service to 100 public and nonprofit sites, yet connections depended on the commercial residential build-out, and service reached only 28 of the 100 sites as of 2019 (Stratton et al. 2022). Public benefits that ride on commercial priorities inherit commercial timelines. Chattanooga, by contrast, supplies the counter-model: HCS EdConnect gives free 100 Mbps service to families of roughly 28,500 K-12 students (Community Networks 2021), the essential-public-service premise made concrete.

The labeling rule bears down hardest on the remaining bright spots, which the memo files under descriptive texture: Cary's 41.1% remote-work share in the Research Triangle (SmartAsset 2023), Austin's projected 42% STEM job growth for 2014 to 2024 (GovTech 2020), and the Triangle's early field trial of GFiber Labs' 20 Gbps residential service (GFiber Labs 2023). Suggestive, correlational, and labeled as such.

A Playbook With One Cautionary Tale

Outcomes, measured or missed, ride on deployment choices, and Google Fiber's strategic arc produced three tested entry models. The 2010 announcement drew over 1,100 municipal applications; Kansas City won in 2011, Austin followed in 2013, the Research Triangle in 2015. By 2016 the company paused new deployments under cost pressure, then resumed expansion in 2022 around multi-gig tiers and public-private structures; as of early 2024 GFiber advertises symmetric 1, 2, 5, and 8 Gbps plans for approximately $70 to $150 per month.

Provo demonstrates municipal acquisition: Google bought the city's financially struggling iProvo network for a symbolic $1 and committed to gigabit upgrades, a template for rescuing stranded public infrastructure (BBCmag 2022). West Des Moines demonstrates public-private conduit leasing: the city built and owns a conduit network, leases it to private ISPs, and landed Google Fiber as the anchor tenant, de-risking the private investment and accelerating deployment (Fiber Broadband Association 2022). Louisville demonstrates the failure mode: experimental nanotrenching left failing sealant and exposed cables, and Google abandoned the market entirely in 2019 (CNET 2024). Physical infrastructure punishes unproven construction methods.

Even the coverage map holds the evidentiary line: I generated it in R from Google Fiber's public market lists and press releases, a count of company announcements, 33 metro areas live, 14 on the public wait-list, July 2025.

What BEAD Should Enforce

The catalyst metaphor has since migrated from pitch to policy. The memo points this record at BEAD, the Broadband Equity, Access, and Deployment program, which frames universal, affordable high-speed internet as a catalyst for small business growth and economic opportunity (The White House 2024). Three recommendations follow, each anchored to a market above. First, prioritize open-access conduit policies; West Des Moines shows city-owned conduit lowering entry barriers. Second, mandate symmetrical speed tiers; even the threat of a new entrant pulls incumbent speeds upward. Third, tie funding to enforceable low-income adoption milestones; Kansas City and Austin show availability arriving well before equitable adoption. The memo phrases the goal as productivity gains that diffuse inclusively.

The summary table hands policymakers one further lesson, teaching as much through its sparse cells as its filled ones. Macroeconomic impact stands measured for Chattanooga alone, housing premia exist for two states and zero pioneer metros, and the headline job figures carry asterisks marking model-derived estimates. I read that sparseness as the real finding: cities stake enormous capital, $84 million for one Kansas City phase, while most of the outcome matrix sits empty. Google promised a catalyst; the table records how little of the reaction anyone has measured. Honest synthesis arranges what exists, flags what each number can bear, and marks, cell by empty cell, where the next studies belong.

Boundaries

  • The memo synthesizes published studies and reports zero original causal estimates; every effect size belongs to the cited authors and carries their limits.
  • Causal labels follow the memo's rule: rigorously documented effects count as causal; speed trajectories, remote-work shares, and projected job growth stay descriptive.
  • Chattanooga's dollar and job totals derive from an input-output model per the memo's own asterisk: gross estimated flows rather than net employment change.
  • The coverage map, which I generated in R from hand-collected city lists (Google Fiber site and press releases, July 2025), counts company announcements rather than audited service areas.
  • The BEAD recommendations remain reasoned proposals, and the memo remains a working policy document rather than a peer-reviewed publication; every figure stays checkable against the hosted PDF and the studies below.

Sources