AP1000
Plant Vogtle Units 3 & 4

It Was Well Designed.It Still Didn't Work.

The most experienced reactor company in American history designed it. Federal regulators reviewed it for years. A fixed-price contract was supposed to absorb the risk. The project finished roughly $21 billion over budget and about seven years late.

Six documented failures at Vogtle — and what each one reveals about the way retirement money is usually managed.

Start with the six failuresSchedule a Conversation

For many of the people I work with, Plant Vogtle Units 3 and 4 are not a case study. They are a job site, a career, a decade of working life. Some of you helped build them. So let me say the obvious thing first: the units got built, they work, and they will produce carbon-free power for Georgia for the next sixty to eighty years. That is a real accomplishment by real people, and nothing on this page takes it away.

But the way they got built is one of the most instructive stories in modern American engineering — and almost none of the lessons are about nuclear power.

Vogtle 3 and 4 were designed by Westinghouse. The AP1000 was reviewed by the Nuclear Regulatory Commission for years before a shovel went in the ground. The construction plan was modeled, contracted, insured and guaranteed. Every major decision along the way was a defensible decision, made by qualified people, with the best information available at the time.

And the project still finished at roughly $35 billion against an original estimate of about $14 billion, seven years behind schedule.

That gap is not a story about incompetence. It is a story about the distance between a plan that is well designed and a plan that works. I have never seen a better illustration of what happens to retirement strategies.

$14B → $35B
Original estimate against final cost, including financing
2016 → 2023
Unit 3's promised in-service date against the actual one
14 → 1
AP1000 reactors the cost model assumed, against projects finished
13,000
Technicians who had to be trained because the workforce no longer existed

Figures drawn from Georgia Public Service Commission construction monitoring testimony, Department of Energy review, and Southern Company regulatory filings. Sources listed at the foot of this page.

Six failures

Every one of these was a good idea when it was adopted

That is the point. None of what follows was negligence, and none of it was obvious in advance. Each was a sound decision that met reality and lost. Read the right-hand column of each pair as the part nobody planned for.

Failure 01 — Modular construction

The packaged solution is only as good as the factory

Don't build a reactor stick by stick in the weather. Build it in pieces, indoors, in a purpose-built plant at the Port of Lake Charles, then ship the modules to Georgia.

As designed

Controlled factory conditions produce controlled quality. Repetition drives cost down across a fleet of reactors. Weather, congestion and site labor variability all come off the critical path.

It was — and still is — a genuinely good idea.

As built

The independent monitor reporting to Georgia regulators concluded in late 2012 that the fabricator clearly lacked nuclear industry experience and was not prepared for the rigor the work demanded. Welds of the wrong type had to be found and repaired.

Module deliveries slipped. Federal inspectors were still documenting weld nonconformances at the module shop years later. When the anticipated pipeline of reactor orders never materialized, the modular facility itself was cancelled.

The retirement version

Target-date funds. Model portfolios. The default allocation sitting inside your 401(k) right now. Standardization is legitimate engineering, not a scam — but a factory product is built for the average unit.

It does not know that forty percent of your net worth sits in your employer's stock, that you have a pension election in ninety days, or that you are leaving at 57 and need a bridge to Social Security. The package is only as good as the assumptions that went into the factory, and nobody at the factory has met you.

Failure 02 — The fixed-price contract

A guarantee is a promise from a balance sheet

Transfer the risk. The 2008 engineering, procurement and construction agreement was fixed-price: the contractor, backed by its parent company, would absorb overruns and delays.

As designed

A 2015 settlement reaffirmed the arrangement. In January 2016 the owners received $920 million in letters of credit. A parent guarantee stood behind the contractor's payment obligations, including liability for abandonment of the work.

On paper, the owners had moved the danger off their own books.

As built

In February 2017 the parent wrote off more than $6 billion and announced it was exiting new plant construction. Six weeks later, on March 29, the contractor filed for Chapter 11. The fixed-price agreement was rejected in bankruptcy.

The guarantee's collectability came into serious question as the guarantor's own going-concern status was raised. One co-owner's budget moved from $5 billion under the "fixed price" agreement to a preliminary range of $6.5 to $7.3 billion.

The retirement version

And read the caps. Even had everything performed exactly as written, liquidated damages were capped at roughly 10% of the contract price and abandonment liability at 40%. The guarantee was never sized to the actual exposure. It never is.

Annuity income riders. Buffered and structured products. Pension obligations. Long-term care contracts. These can be entirely appropriate tools — I use some of them. But "guaranteed" is not a property of the product. It is a statement about a counterparty's ability and willingness to pay.

Ask three questions before you lean on one: Who is on the other side? Where is the cap? What happens to me if they are not there?

Eight questions, about two minutes
Do you know where your guarantees actually cap out?

The Portfolio Preparedness Review walks through the parts of a retirement strategy most likely to behave differently than the brochure suggests.

Take the review
Failure 03 — The basemat rebar

The plan on paper is not the plan as built

The nuclear island basemat — the foundation the entire plant sits on — was fully specified in a design the federal government had certified. Bar size, spacing, embedment, tolerances. Nothing ambiguous.

As designed

A certified design, reviewed and approved before construction, with every reinforcement detail fixed in the design control document. The most scrutinized foundation drawing in American civil engineering.

As built

In April 2012 inspectors found that the rebar as actually installed did not match the approved design. Work stopped. The license amendment reconciling the as-built condition was not approved until October.

Roughly six months lost on the most foundational element of the project, before a yard of structural concrete could be poured.

The retirement version

The design was not wrong. The building did not match it.

The investment policy statement says 60/40; the accounts have drifted to 78/22 because nobody rebalanced through a long bull market. The beneficiary designation still names a former spouse. The Roth conversion has been "part of the plan" for four years and has never been executed. The old employer's 401(k) from 2009 is still sitting in a money market fund.

Design conformance is a habit, not a document — and the longer a deviation stays buried, the more expensive it is to correct.

Failure 04 — The cable raceways

Redundancy that shares a path is not redundancy

Safety-related electrical cables are routed through separated raceway systems so that no single event — a fire, a flood, an impact — can disable both redundant paths at once.

As designed

Physical separation between redundant trains. The entire concept of defense in depth rests on it: one failure can never propagate to take out the backup that exists to cover it.

Elegant, well understood, and non-negotiable under federal regulation.

As built

Quality problems surfaced in Unit 3's raceway system in late 2020 and were deemed significant in early 2021. A federal special inspection that June found safety and non-safety cables had not been adequately separated for reactor coolant pumps and equipment needed to safely shut the reactor down.

Inspectors also found instances where construction quality issues were known but never entered into the corrective action program. Regulatory oversight was increased and fuel load was gated on remediation.

The retirement version

This is concentration risk, and it is the single most common structural failure I see in utility households. The paycheck comes from one company. The pension comes from the same company. The retiree medical comes from the same company. The 401(k) is heavy in that company's stock. And the "diversified" index fund sitting alongside it has roughly a third of its weight in ten names.

Four separate accounts, four separate statements, one point of failure.

The second half of that inspection finding matters just as much: people knew, and did not write it down. The retirement equivalent is not opening the statement.

A two-minute system check
Where does your retirement have a single point of failure?

The Retirement Grid walks six connected areas of a plan and shows you where redundancy exists only on paper — the same question the raceway inspection was asking.

Run the Retirement Grid
Failure 05 — The vibrating pipe

Some failures only appear under load

Cold hydro testing, hot functional testing, startup testing — an enormous, expensive, deliberate regime whose entire purpose is to find problems before the plant carries real load.

As designed

Test everything before it matters. Find the defects while the reactor is still cold, while the consequences are schedule and money rather than safety.

This one worked exactly as intended — and that is precisely the lesson.

As built

Fuel was loaded into Unit 3 in October 2022. In January 2023, during startup and pre-operational testing, vibration was detected in piping associated with the automatic depressurization system. Not a safety event. The remedy was mechanically simple — added support, metal plates inserted into struts.

But the schedule carried a cost of roughly $15 million per month. Unit 3's in-service date slid from the first quarter to April, then to May or June 2023. In February 2024, Unit 4 — built by a seasoned team with everything learned from Unit 3 — hit its own cooling-system vibration issue and slipped into the second quarter.

The retirement version

This is sequence-of-returns risk, and it is the reason I do this work. A portfolio that behaved beautifully across thirty years of contributions can fail in the first five years of withdrawals, because selling into a drawdown is a structurally different operation from buying into one.

The failure mode is invisible during accumulation. It only appears under load — and by the time it appears, the cheap fixes are behind you.

Note Unit 4 carefully. Experience does not automatically transfer. Having done something once does not immunize you against the same class of problem the second time.

Failure 06 — The learning curve

Your plan's math assumes a world

First-of-a-kind projects are expensive; nth-of-a-kind projects are not. Fourteen AP1000 reactors were envisioned, and the enormous fixed costs would spread across a fleet.

As designed

Engineering, licensing, supply chain and a trained workforce are one-time investments. Amortize them across fourteen units and the marginal reactor becomes competitive with anything else on the grid.

The arithmetic was sound. Unit fourteen would have been cheap.

As built

Natural gas prices fell. Demand forecasts fell. Of the fourteen, only Vogtle was carried to completion. The workforce did not exist and had to be created. The supply chain had to be rebuilt from nothing.

The design was not complete when construction began, and changes had to be processed as license amendments the entire way through. The Department of Energy's review of the overruns cited incomplete design, inadequate quality assurance, limited constructability, and a shortage of experienced labor.

The retirement version

The engineering is not what failed. The assumed world is what failed.

Every Monte Carlo simulation, every 4% rule, every glide path is calibrated against a historical world: a certain range of inflation, a certain correlation between stocks and bonds, a certain cost of capital, a certain level of sovereign debt. Those are not laws of physics. They are the retirement-planning equivalent of "we expect fourteen orders."

Where I'll be direct with you

I will say plainly what I believe, because I have been writing about it for years. We are living through monetary and fiscal strain unlike anything in most of my clients' working lives.

$18T → $39T
U.S. federal debt, roughly one decade apart

I do not know the timing, and I do not trust anyone who tells you they do.

But a plan whose safety rests on the next thirty years resembling the last thirty is making the same bet the reactor business made about its order book.

The record

Fifteen years, in the order it happened

Nothing here was a single catastrophic decision. It accumulated — which is exactly how retirement plans come apart.

  • 2008
    Fixed-price EPC contracts signed. Contractor and parent bear the overrun risk.
  • February 2012
    Combined licenses issued. First nuclear concrete placed. Unit 3 targeted for 2016.
  • April – October 2012
    Basemat rebar found not to match the certified design. Roughly six months lost.
  • December 2012
    Independent monitor reports the module fabricator lacked nuclear experience. Wrong-type welds repaired.
  • January 2016
    $920 million in letters of credit delivered to the owners. The risk transfer looks intact.
  • February – March 2017
    Parent writes off more than $6 billion. Contractor files Chapter 11. The fixed-price contract is rejected in bankruptcy.
  • 2020 – 2021
    Electrical raceway quality problems surface. Federal special inspection finds inadequate cable separation and unreported conditions. Oversight increased.
  • October 2022
    Unit 3 fuel load completed after remediation.
  • January 2023
    Vibration found in cooling-system piping during startup testing. Roughly $15 million per month of delay.
  • July 2023
    Unit 3 enters commercial operation — seven years after the original date.
  • February – April 2024
    Unit 4 hits its own cooling-system vibration issue, then enters commercial operation.
What actually worked

They got there — not because the plan was right, but because somebody was inspecting it

Vogtle 3 and 4 are running. Unit 3 entered commercial operation in July 2023, Unit 4 in April 2024. Together they will supply Georgia with carbon-free power for decades.

They got there despite a plan that was wrong repeatedly, because there was continuous, independent, adversarial inspection. Problems were surfaced, priced and corrected while correction was still possible.

Federal inspectors were on site throughout. The Georgia Public Service Commission retained its own independent construction monitor — a nuclear professional with more than fifty years of experience who testified semiannually for seven years and reported to the regulator, not to the builder. His job was to say what was actually happening, whether or not anyone wanted to hear it.

That is the model worth copying. Not a better forecast. A better inspection regime.

You will not get a retirement plan right the first time. Nobody does. What separates the plans that work from the plans that fail is not the quality of the original forecast — it is whether somebody is inspecting the thing honestly, on a schedule, with no incentive to tell you it's fine.
The question nobody asks

Vogtle had inspectors. What inspects the firm holding your retirement?

The reason those two reactors are running is not that the plan was right. It is that somebody was always looking — and that somebody did not work for the people being looked at. That is a structural feature, deliberately built and expensive to maintain.

Most retirement relationships do not have it.

Plant Vogtle Units 3 & 4
Federal inspectors resident on site, continuously, for the life of construction
An independent construction monitor retained by the Georgia Public Service Commission — reporting to the regulator, not the builder
Sworn public testimony twice a year, for seven years
A corrective action program that problems were required to be entered into
Findings published where ratepayers could read them
A typical retirement relationship
Supervision performed by the firm that employs the advisor
Inspection cycles and methods largely elected by the firm within the rules
Reviews that are internal, and not published
Disclosure that genuinely exists — in documents most clients never open
One party with a real incentive to look: you

None of that is an accusation, and none of it is evidence that anything is wrong. Self-supervision is the standard architecture of this industry. It is lawful, it is disclosed, and most of the people working inside it are conscientious.

But notice what it means. The inspection regime that saved Vogtle from its own plan was adversarial by design — somebody was paid to go looking for what the builder had missed. Yours is not built that way. Which puts the inspection back on you.

Run the inspection yourself

It takes about twenty minutes, it costs nothing, and you do not need anyone's permission.

01 — The broker record

brokercheck.finra.org — look up both the firm and the individual, then open the Disclosures tab. Reportable events include customer complaints, arbitrations, regulatory actions, terminations and certain financial events.

02 — The adviser record

adviserinfo.sec.gov — the Investment Adviser Public Disclosure database. It draws on a different registry than BrokerCheck, so check both; older brokerage-side events sometimes appear in one and not the other.

03 — The brochure

From that same record, download Form ADV Part 2A. Item 9 is Disciplinary Information. Item 5 is Fees and Compensation — including any payments received from product providers. Part 2B covers the individual actually advising you.

04 — Form CRS

Two pages, required of firms serving retail investors. Item 4 is Disciplinary History, and it asks the question on your behalf: do you or your financial professionals have legal or disciplinary history? Investor.gov/CRS has a free search tool.

A disclosure is not proof of wrongdoing. Some are decades old. Some were resolved in the professional's favor. Some reflect a complaint that went nowhere. The point is not to go hunting for a villain — it is to know what you are looking at before you rely on it, and to ask for an explanation of anything you find, in writing.

Vogtle's monitor had one job: to say what was actually happening, whether or not anyone wanted to hear it. Until someone in your financial life holds that job, the job is yours.

Run it yourself

Your own construction monitoring report

Five questions. If you can answer all five without opening a statement, your plan is in better shape than most.

01 — Design conformance

Where does my plan on paper not match how my accounts are actually built right now?

02 — Guarantees

What in my plan is guaranteed, who is guaranteeing it, and where is the cap written?

03 — Shared paths

Which of my "diversified" holdings depend on the same single point of failure — my employer, one sector, ten stocks?

04 — Load testing

What has never been tested under load? Specifically: what happens to my withdrawal plan if the market falls 30% in my second year of retirement?

05 — Assumed world

What does my plan assume about inflation, interest rates and the next three decades — and how much of my outcome depends on those assumptions holding?

Most people cannot answer the fourth one. That is not a character flaw. It is the same reason the vibrating pipe was not found until startup testing: some things simply do not reveal themselves until you run the system for real — and by then you want somebody watching who is paid to notice.

Take the Portfolio Preparedness ReviewHow I think about risk

An independent look at how your plan is actually built — not how it was designed

I work with Southern Company, Georgia Power and utility-industry employees and retirees on concentration risk, pension elections and sequence-of-returns risk. Bailey Financial Services is a fee-only fiduciary. No products, no commissions, no one else's shelf to fill.

Wilder Bailey
Founder & Principal
Watkinsville, Georgia
Wilder@BaileyFS.net
Schedule a ConversationMeet WilderWhat is a fiduciary?

If this was useful, read these next

Companion piece
One Stock
Five nuclear events that rewrote the rules nobody drilled for — mapped to the utility career built on a single stock.
Reference
Someone Who Sat Where You Sit
Forty years at Southern Nuclear, retired at Plant Vogtle, a client for twenty years. His words, his name, his permission.
The mechanics
Concentrated Risk
You may own the same stock four different ways without knowing it. How the overlap hides.

Sources

  1. Don Grace, PE — Plant Vogtle Construction Monitor, 2017–2024 — "What Was Learned from Building New Nuclear Reactors?" POWER, April 2025.
  2. Center on Global Energy Policy, Columbia University SIPA — "Vogtle Unit 3 Has Started Commercial Operations. What's Next for the AP1000?" (summarizing the Department of Energy review of root causes).
  3. U.S. Nuclear Regulatory Commission — special inspection findings and enforcement correspondence on the Vogtle Units 3 and 4 electrical cable raceway system, June–November 2021.
  4. Taxpayers for Common Sense — "DOE Loan Guarantee Program: Vogtle Reactors 3 & 4," March 2019 (basemat rebar chronology and module fabrication findings).
  5. Congressional Research Service — "Westinghouse Bankruptcy Filing Could Put New U.S. Nuclear Projects at Risk," 2017.
  6. Southern Company and Oglethorpe Power SEC filings, 2017 and 2023 — EPC contract terms, liability caps, guarantee obligations and schedule revisions.
  7. POWER, Power Engineering, World Nuclear News and Associated Press reporting on Unit 3 and Unit 4 startup testing and vibration remediation, 2023–2024.

Bailey Financial Services, Inc. is a state-registered investment adviser. This material is for educational purposes only and is not investment, tax or legal advice. It does not constitute a recommendation regarding any security, product or strategy, and it is not an offer to buy or sell any security. Investing involves risk, including possible loss of principal. Past performance does not guarantee future results. References to Plant Vogtle, Georgia Power and Southern Company are drawn from public regulatory filings and press reporting and are used for illustration only; nothing on this page should be read as a statement about, or on behalf of, any company or its securities, or as a recommendation to buy or sell any company's stock.