Trump Ordered $200 Billion of Mortgage Bonds. The PSPA Only Has Room for $113 Billion. Somebody Has to Sign Something.

If you're new here: I'm Glen Bradford. I'm long Fannie Mae and Freddie Mac junior preferred shares and have written the full Fanniegate thesis for years. This post is about a piece of arithmetic that, as far as I can tell, nobody has done in public on the right basis: the President ordered the companies to buy $200 billion of mortgage bonds in January, and the contract that governs the companies — the Senior Preferred Stock Purchase Agreement with Treasury, the PSPA — never had room for it. Every number below is from the companies' own SEC filings and monthly summaries, linked at the bottom. Check my math.

The order

On January 8, 2026, President Trump posted on Truth Social that he was "instructing my Representatives to BUY $200 BILLION DOLLARS IN MORTGAGE BONDS," and that "This will drive Mortgage Rates DOWN, monthly payments DOWN, and make the cost of owning a home more affordable" (Scotsman Guide, CNBC). FHFA Director Bill Pulte confirmed on X that "Fannie and Freddie are the entities that will do the purchases," and, "We are on it, Mr. President!" The next day Treasury Secretary Bessent told reporters the goal was to roughly match the Fed's runoff: "the Fed has about $15 billion of roll-off every month… the idea is to roughly match the Fed, which has been pushing the other way."

Four days later, on January 12, FHFA emailed the companies and lifted its own limit on how much agency MBS each could hold for investment — from $40 billion to $225 billion apiece, "effective immediately" (AP via Fortune). Pulte's framing when that leaked: FHFA "simply gave each entity legal flexibility to go beyond their previous caps," and the companies would not "exceed $200 billion."

Note what that number is. $225 billion is not a number FHFA picked. It is the ceiling in the PSPA. FHFA raised its own sub-limit all the way up to Treasury's ceiling — and stopped, because that is where FHFA's authority stops.

The cap

Section 5.7 of the PSPA, as amended by the January 14, 2021 letter agreement, in full:

"Seller shall not own, as of any applicable date, Mortgage Assets, in excess of (i) on December 31, 2020, $250 billion (ii) on December 31, 2021, $250 billion, or (iii) on December 31, 2022, and thereafter, $225 billion. For purposes of this section 5.7, interest-only securities shall be measured at 10 percent of their notional value, and other Mortgage Asset balances or amounts shall be measured at unpaid principal balance." (letter agreement, EDGAR)

That is a covenant between each company and the United States Treasury. FHFA cannot waive it. Congress is not involved in it. It can be changed exactly one way: a letter agreement signed by Treasury and by the company "acting through the Federal Housing Finance Agency as its conservator" — the same instrument used in 2012, 2017, 2019, 2021 (twice), and most recently on January 2, 2025 (Fannie 8-K exhibit). Every one of them landed as an 8-K, and I checked every 8-K both companies have filed since: not one has been signed in 2026. Both companies recite the covenant in their current 10-Qs. Fannie: "The amount of mortgage assets that we may own in our retained mortgage portfolio is capped at $225 billion under the terms of our senior preferred stock purchase agreement with the U.S. Department of Treasury." Freddie: "The Purchase Agreement limits the size of our mortgage-related investments portfolio to a maximum amount of $225 billion."

Now the arithmetic.

The math, on the cap's own basis

Here is where the two companies stood against the cap, on the cap's own measurement basis (unpaid principal plus 10% of interest-only notional), from their Q2 2026 10-Qs:

Cap basis, $ billions	Dec 31, 2025	Jun 30, 2026	Cap	Room left
Fannie Mae	$134.2	$177.5	$225	$47.5
Freddie Mac	$161.2	$159.8	$225	$65.2
Combined	$295.4	$337.3	$450	$112.7

Fannie's figure includes $2.4 billion for interest-only notional. Freddie's includes $21.2 billion — Freddie holds a large book of IO strips, and one-tenth of their notional counts against the cap. That single line is why the public estimates of how much room the companies had ran about $24 billion high.

Read the top row of that table again. On the day the President gave the order, the two companies had $154.6 billion of combined room under the PSPA. The order was for $200 billion. The order never fit. Not after six months of buying — on day one.

Why did nobody say so? Plenty of people discussed the caps — J.P. Morgan and Bank of America told National Mortgage News in May that both companies were "likely to hit their portfolio caps in September" (link); Rithm Capital wrote in June that "any acceleration that approaches cap limits would require either an amendment to the PSPA framework or a structural policy decision" (link). But the number everyone used for the room was wrong. The widely-circulated capacity estimate in January — Loomis Sayles put the combined portfolios around $247 billion as of November against the $450 billion combined cap and concluded there was roughly $200 billion of remaining capacity — used headline portfolio balances and skipped the IO add-on, which was about $24 billion at year-end. KBW told clients in June both portfolios were "potentially reaching the $250B cap by year-end" (National Mortgage News) — a cap that hasn't existed since December 31, 2022. The headline monthly portfolio numbers everyone quotes don't include that line. The 10-Qs do. So the Street's math said the program fit comfortably, and it didn't.

What actually happened: they bought, then stopped

From the companies' own monthly summaries (Fannie, Freddie), net change in agency securities held, $ millions:

2026	Jan	Feb	Mar	Apr	May	Jun	H1
Fannie	+11,524	+4,224	+17,662	+5,535	−5,367	−615	+32,963
Freddie	+3,970	+7,038	+700	−105	+117	−70	+11,650
Combined	+15,494	+11,262	+18,362	+5,430	−5,250	−685	+44,613

First quarter: +$45.1 billion. Second quarter: −$0.5 billion. Fannie was a net seller of agency MBS in May. Freddie has been flat since March. Against a $200 billion order and Bessent's stated ~$15 billion-a-month pace, the companies did $44.6 billion in six months — about $7.4 billion a month — and then went to zero. On February 25, Pulte told Fox Business the President had deployed "roughly $200 billion" and that "$200 billion reduced mortgage rates. Boom right away" (Fox Business). At the end of February the actual number was about $27 billion.

Fannie's explanation on its Q2 earnings call: "Growth in our agency MBS and lender liquidity portfolio slowed in the second quarter versus the last three quarters. We continue to monitor market dynamics, spreads, and returns to make sound decisions about balance sheet deployment while remaining in compliance with portfolio limits and managing interest rate risk." Fair enough — spreads compressed after the March buying and the relative value thinned out; as Rithm Capital put it, "the GSE bid is spread-conditioned". But notice the last clause. And notice that Freddie's cap-basis balance actually fell in the first half — it added $11.6 billion of MBS while running off $16 billion of unsecuritized multifamily loans. Freddie made room by shrinking something else. That is what managing to a ceiling looks like.

Neither company has ever named the "$200 billion" program in an SEC filing. The phrase appears zero times in Fannie's 10-K, its Q1 and Q2 10-Qs, or Freddie's. What the filings describe is "FHFA's January 2026 increase to our agency MBS investment limit" and the $225 billion cap. National Mortgage News put it politely in June: "policymakers have not been fully transparent in disclosing how this is being managed or will end" (link).

So what has to happen

Add it up. $200 billion ordered. About $45 billion done. About $155 billion left. About $113 billion of room, and that room is also where delinquent-loan buyouts and lender-liquidity loans live — Fannie's retained loans grew $9.6 billion in the first half on their own. There are exactly three ways this resolves:

1. The order quietly dies. The companies bought what made sense at March spreads, the rhetoric moves on, and "$200 billion" becomes a number nobody mentions again. Possible. It would mean the President's most concrete housing-affordability action of 2026 was executed at 22%.
2. The companies buy to the wall. Roughly $113 billion more, cap-basis, and then a hard stop. Barclays' MBS desk flagged this in March: "Absent an increase in PSPA caps the market will need to price in the eventual GSE exit sometime this summer. That's likely to be the defining driver of MBS performance for the remainder of 2026" (NMN). Bank of America's Jeana Curro in May: "Enjoy it while it lasts, and then I think it is over. I don't think we're seeing more buying after that." Summer and September slipped because the buyers slowed down, not because the wall moved.
3. Treasury and FHFA sign a letter agreement raising the cap. KBW's Bose George said as much in January: "since the PSPA is an agreement between FHFA and the U.S. Treasury, it could be amended to allow the GSEs to purchase more MBS, either temporarily or permanently" (NMN). Community lenders formally asked Bessent and Pulte for a $300 billion-each ceiling last October (Scotsman Guide).

Option 3 is the only one where the President's order gets filled. And option 3 means Bessent's signature on the PSPA.

Why that signature is the whole ballgame

If you have read this blog for more than a week you know why that matters. The PSPA is not a portfolio-cap document that happens to mention the senior preferred. It is the senior preferred. It is the document under which Treasury holds its $193.5 billion of senior preferred stock — with a liquidation preference that reached $380.8 billion on June 30 and steps to $388.4 billion on September 30, because every dollar of net worth the companies earn is added to it the following quarter — and its warrants for 79.9% of the common, and it carries the covenants that have kept both companies in conservatorship for eighteen years. The companies' own Q2 capital disclosures showed that the senior preferred is the only reason their regulatory capital reads negative. Every path out of conservatorship runs through amending it.

For a year the market has been waiting for that amendment as a discretionary policy choice — Bessent and Pulte will get to it when the offering is ready, when spreads cooperate, when the calendar allows. The portfolio cap changes the character of the decision. It is now a decision with a deadline attached to somebody else's promise. Either the President's $200 billion is a real number, in which case Treasury has to open the PSPA to make room for it, or it isn't. And once the document is open for one clause, the question of what else gets written in it is live. The last two amendments, January 2021 and January 2025, each rewrote several sections at once.

I want to be careful here. A cap raise can be a one-paragraph letter that touches nothing else — that is what the bear case looks like, and it is entirely possible. But consider what the last amendment did. The January 2, 2025 letter agreement — signed by Yellen's Treasury and Sandra Thompson's FHFA in the final weeks of the prior administration — made Treasury's written consent mandatory for any exit from conservatorship, and its side letter requires FHFA, before asking for that consent, to deliver a proposal addressing "any amendments to the SPSPA, senior preferred stock certificate or warrant that may be required to implement the recommended approach" (Treasury release). The next amendment is contemplated by the last one. It will be the first of Bessent's Treasury. And it will be signed by the same two people who spent 2026 saying in public that the exit is coming — Pulte as recently as February, when he said the companies were "definitely ready" for a 2.5–5% sale and that timing was "up to the president" (NMN); the President in June, when he said "It's not a rush."

The second clock: September 8

There is another government decision sitting in the same window, and it is not discretionary either.

On July 24, 2026, the D.C. Circuit affirmed Judge Lamberth's judgment in the shareholder class action — Fairholme Funds v. FHFA, No. 25-5113 — in full (slip opinion). Senior Judge Ginsburg wrote for a unanimous panel with Judges Walker and Childs; the panel I previewed and recapped in April. The disposition, in one sentence: "For the foregoing reasons, the judgment of the district court is Affirmed." That is $612.4 million in jury damages plus $199.65 million of prejudgment interest — a final judgment of $812.05 million, accruing post-judgment interest at 5.01% since March 2024, so roughly $914 million today — for breach of the implied covenant of good faith when the government imposed the net worth sweep in 2012. The judgment runs jointly and severally against each company and FHFA as conservator. (The panel also rejected the Berkley plaintiffs' $48 billion restitution cross-appeal, so nobody should read this as an unwind. It is a damages verdict, affirmed.)

The government has 45 days to seek rehearing en banc. Forty-five days from July 24 is Monday, September 7 — Labor Day — so the deadline rolls to Tuesday, September 8, 2026. If no petition is filed, the mandate issues around September 15, and the 90-day window to petition the Supreme Court runs to October 22 — at which point, under the court-approved plan of allocation, the judgment is "final and non-appealable" without anyone in Washington lifting a finger. Both 10-Qs use identical non-committal language: Fannie says "Consideration is being given to the possibility of further action"; Freddie says "The defendants are evaluating potential next steps." No statement from FHFA or Treasury.

To be precise about what September 8 is and isn't: it is not the payment date. Under the plan of allocation, distribution follows finality — an allocation notice to the classes, court approval, a record date, disbursement through brokers — and how quickly that runs after October 22 depends on how much of it is teed up in advance. Per-share recoveries will be modest relative to par; this is a damages verdict, not a recap. What September 8 is: the day FHFA, under Bill Pulte, decides whether to keep litigating against the shareholders of the companies it says it wants to take public, or to let a unanimous appellate ruling that the net worth sweep breached the shareholders' contracts become settled law.

Two decisions, one calendar

Put them next to each other.

• By September 8, FHFA decides whether the net-worth-sweep breach verdict becomes final.
• Sometime between now and roughly Q1 2027 — sooner if buying resumes at anything like the Q1 pace, later if the companies stay parked — Treasury and FHFA either amend the PSPA's portfolio cap or concede the President's $200 billion order won't be filled.

Both decisions belong to Bessent and Pulte — Pulte now also serving as acting Director of National Intelligence, which is its own comment on FHFA bandwidth. Neither requires Congress. Both involve the government choosing its posture toward the pre-conservatorship shareholders and toward the document that keeps the conservatorship in place. If the September 8 deadline passes quietly and a letter agreement follows, that is a government that has stopped fighting the old war and started clearing the paperwork for the new one. If FHFA petitions for rehearing and Treasury lets the MBS program die at 22%, that is a government still running out the clock. Either way, we will know a great deal more about timing by mid-September than we did in July. Add the September 30 step-up in the liquidation preference — another $7.6 billion added to a number Treasury will eventually have to resolve — and the fall of 2026 is a lot busier than the summer's silence suggests.

What I'm doing

Nothing new. Long the junior preferred, same as I've been for years. The portfolio cap doesn't change the position; it puts a date on a decision I've been expecting for eighteen years. I'll be watching for a rehearing petition on the D.C. Circuit docket through September 8, the July monthly summaries around August 26–28 to see whether the buying resumed, and any letter agreement on EDGAR — the prior ones showed up as 8-K exhibits within days of signing.

---

The Sources

• Fannie Mae Form 10-Q, Q2 2026 — retained portfolio table (p. 14), $225 billion cap language, "for this purpose" $177.5 billion, Lamberth accrual and appeal status
• Fannie Mae Form 10-Q, Q1 2026 — the $36.3 billion Q1 jump and the January 2026 FHFA limit increase
• Fannie Mae Form 10-K, 2025 — October 2025 $40 billion FHFA limit; January 2026 increase "provided that our total mortgage assets do not exceed the $225 billion cap"
• Freddie Mac Form 10-Q, Q2 2026 — Table 9, mortgage-related investments portfolio "for purposes of Purchase Agreement cap" $159,750 million, IO add-on $21,179 million
• Fannie Mae Monthly Summary, June 2026 · Freddie Mac Monthly Volume Summary, June 2026 — month-by-month agency securities balances
• January 14, 2021 letter agreement — Section 5.7 cap text · FHFA/Treasury September 14, 2021 suspension — the cap was not among the suspended provisions
• Scotsman Guide, Jan 2026 · CNBC, Jan 8 · Fortune/AP, Jan 24 · Fox Business, Feb 25
• National Mortgage News, Mar 28 (Barclays) · Jan 5 (KBW) · Jun 1 · Jun 29 · Loomis Sayles, Jan 22 · Fannie Q2 earnings call transcript
• D.C. Circuit opinion, Fairholme Funds v. FHFA, No. 25-5113 (July 24, 2026) · Final Judgment, Mar 20, 2024 · Order Governing Plan of Allocation · FRAP 40 · FRAP 41

If You Want to Go Deeper

• Full Fanniegate thesis — the whole story
• The Q2 2026 ERCF Capital Disclosures — why the senior preferred is the only thing standing between the companies and positive regulatory capital
• The 2026 DFAST stress test — the balance sheet doesn't flinch at a worse-than-2008 crash
• The Trump GSE Promotion Timeline — the policy drumbeat, dated and sourced
• D.C. Circuit oral argument recap, April 21 — the hearing that produced the July 24 affirmance
• Fanniegate Timeline — interactive milestone history

---

I hold long positions in Fannie Mae and Freddie Mac junior preferred shares. This post is my personal opinion and is not financial advice. All portfolio and cap figures are from the companies' SEC filings and monthly summaries linked above; the "room left" arithmetic and the reading of what it implies are mine. Whether Treasury amends the PSPA, when, and what else it writes in when it does are open questions — this post argues they are now questions with a clock on them, not that the answer is known. Do your own research. The full thesis is at glenbradford.com/fanniegate.