Six Earnings Updates!

🌾 Welcome to StableBread's Newsletter!

Last Sunday I released SECSift. It's an EDGAR reader that turns SEC filings into clean, readable documents. It lets you search every word, export anything, merge tables, highlight what matters and gray out the rest, and compare against previous filings.

It's completely replaced EDGAR for me and a few of my equity research colleagues!

But there’s a small learning curve, so I made this video walking through the entire reader:

You can also just click through this 3-minute interactive demo (only viewable on your computer).

As a reminder, annual plans (for more AI usage, storage options, exporting) are 25% off with code “SB25” through August 31.

It’s been a packed earnings week! This post breaks down my quarterly analysis for six names from the write-up archive: Toast (Tuesday), Root, Fluence, and Energy Fuels (Wednesday), plus Celsius and Doximity (Thursday).

This is a dense post that I’ve been working on the entire week, so I hope you find it helpful!

Toast (NYSE: TOST) is up over 40% since my June 2nd write-up, where I set my buy target below $25/share.

Here’s what’s notable following the Q2 report:

On the customer side, Best Western named Toast an endorsed food and beverage vendor, which opens the door to the thousands of hotel restaurants it has across the US and Canada. TGI Fridays expanded the partnership and accelerated its UK rollout after the pilot.

Wanted to highlight this quote as well:

Very happy with earnings overall! The one soft spot is the one my write-up called. FCF fell to $130M from $208M, on inventory up $81M in the quarter to $217M, as Toast pre-buys memory ahead of the DRAM and NAND shortage.

Hardware gross profit also widened to -$68M from -$54M, which is not unexpected.

CFO Elena Gomez said Toast has memory supply locked for both 2026 and 2027, has already reduced the expected cost impact for both years versus its original plan, and expects "structurally better hardware margins than before" once the market stabilizes.

The Q3 guide also has adjusted EBITDA at $210-220M, a midpoint below the $221M just posted. That's not a slowdown, it's Toast putting the beat into its newer markets (enterprise, international, and retail), where ARR is on pace to nearly double to $200M this year, plus AI products like Toast IQ.

At $24/share Toast was ~16x forward EV/EBITDA. At $34/share it's ~22x. My $30/share base case was an 18-month target and the stock cleared it in two months!

The $40-44 bull case now requires Toast IQ Grow attach past 20% and memory normalizing, not just the core compounding. On the earnings call, management only shared that Toast IQ Grow is on track to be the fastest Toast product to $10M of ARR and already runs positive margins.

Root (Nasdaq: ROOT) reported $25.4M of net income in Q2, up 15.5% y/y, while its policy count fell by ~11,500 from March.

The net combined ratio came in at 92.1% against 95.2% a year ago, a second straight quarter in the low 90s after Q1's 91.4%. The stock has gone from $60.28 before earnings (same price as my original May 14 write-up) to about $50/share.

Root prices car insurance on smartphone telematics, how you brake, turn, and handle your phone, rather than on demographics. It now cedes just 1.2% of premiums to reinsurers, so underwriting results land on its own book with almost no buffer.

Most of the reaction says Root bought the profit by cutting marketing, and that the loss ratio is deteriorating underneath. The filings show Root moved most of that marketing money rather than cutting it, and that the loss ratio rose because premiums came down, not because claims went up.

$61.3M of H1 net income already beats all of last year's $40.3M. And the one-time items in the quarter net out, so that profit is underlying, not an accounting boost.

Q2 carried ~$10M of favorable prior-year reserve development, worth about 2.8 points, so the underlying number is closer to 94.9%. But the $4.9M refinancing charge and a $4.4M private equity write-down roughly cancel that benefit. Adjusted EBITDA rose 16.5% to $43.8M and annualized ROE was ~31%.

The improvement came from expenses, not underwriting. Of the 3.1 points the combined ratio improved, the expense ratio delivered 3.0 of them, falling to 26.1% from 29.1%. The net loss and LAE ratio (LAE is the cost of settling claims) barely moved, 66.0% against 66.1%.

The expense lines show what moved:

Net the $10.6M against the $7.0M and acquisition spend fell $3.6M, not the $11.8M the marketing line shows. Root moved the money from sales and marketing into other insurance expense, and management calls the direct pullback counter-cyclical, planning to lean back in when competitors retreat.

One warning from the call before extending any of this. CFO Megan Binkley said she "would not run rate the 26% net expense ratio," pointing to Q1 and Q4 (both near 29%) as the more normal base, since part of the G&A drop came from prior accruals reversing.

That reallocation follows where the business is now writing. Partnerships and independent agents produced 51% of new business against 44% a year ago, rising in each of the last three quarters, through Carvana, Hyundai Capital America, Toyota, Experian, and Jerry as of July.

Now the number the market reacted to. The gross accident period loss ratio, the losses from crashes that actually happened in the quarter, rose to 61.6% from 57.4%.

A loss ratio is claims divided by premium, and Root cut the premium. Gross premiums earned fell 0.9% y/y while policies in force grew 6.2%, so each policy now earns ~6.7% less. That decline on its own takes 57.4% to ~61.5%, against the 61.6% reported, with claims costs barely moving.

So the real question is the price under the loss ratio. It reads as the soft cycle plus deliberate segmentation, not lost pricing power, since policies kept growing and claims barely moved while only the price came down.

CEO Alex Timm said Root is still "probably a little overpriced of about 3% or so" even after last year's pricing model lifted customer lifetime values by over 20%, and that modest single-digit rate decreases may keep coming through the book, so premium per policy likely has further to fall.

Premium per policy fell 8.5% y/y to $1,479, and 1.8% from Q1's $1,506. Renewal premium is now 62% of gross earned premium against 55% a year ago, so the book is getting stickier while it gets cheaper.

Neither reading changes what the two numbers multiply into. Premiums in force is policies times premium per policy, and it fell 2.8% y/y to $1.43B after growing 1.9% in Q1, the first decline since the book returned to growth in 2024. It sets earned premium 6-12 months out.

On the call, Binkley said year-end policies in force should be roughly flat versus the end of 2025 if the competitive environment persists. Pair flat policies with a premium per policy that's still being cut, and premiums in force likely stays negative through year-end.

But part of the Root thesis is that it just has to keep expanding, and it is. New Jersey went live in July, state 37 of the 48 targeted by the end of 2027.

Root also repurchased more than $20M of stock in the quarter, the first use of the May authorization.

At $50/share, the market cap is ~$775M on 15.5M shares, or 2.4x June book value of $327.9M. I wrote Root up in May at ~$60/share and 2.9x. Net of the $239.5M held outside the insurance subsidiaries, you're paying ~$535M for the operating company, or 4.4x H1 net income annualized.

The catch is that H2 is the seasonal weak half, so the year lands nearer the low end of the $80-110M base case from May. That $55-80 base case is 2.5-3.5x book value on an 18-22% ROE. The $35-45 bear case is 1.5-2.0x book, where ROE compresses to 12-15% and the soft cycle runs into 2027.

In May, Fluence (Nasdaq: FLNC) told investors its first hyperscaler order would land in the June quarter, and the earnings reported August 5th show ~$850M of data center business booked across the quarter and July.

Yet management cut guidance for the third time since February 2025, and the stock fell ~28.0% before recovering to $13.20/share.

Fluence builds the shipping-container-sized battery systems that utilities and data centers plug into the grid. It doesn't make the battery cells, it buys them and wraps its own engineering, software, and 20-year service contracts around them, which is where its profit comes from.

When I wrote it up three weeks ago at ~$14.50/share, the setup was that the company had no data center revenue at all, so you paid nothing for that possibility, and all the business had to do was hold its usual 11-13% gross margins while the order book compounded.

Start with the order book, which did what management said it would:

Now the margins, which did not hold:

Management attributes the revenue cut to ramp delays at two new contract manufacturing facilities, a three-month delay at the new Houston enclosure plant and quality problems at an international plant that forced rework, pushing ~$90M out of the quarter and ~$400M out of the fiscal year into FY2027.

The margin decline has three named causes in the 10-Q: (1) penalties owed to customers for late projects, (2) cost overruns on the newer Gridstack Pro and Smartstack products, and (3) higher contract costs as battery prices rose.

The miss carried through to the balance sheet. Cash fell to $365.0M against $400M of convertible debt, so a company with net cash a quarter ago now carries ~$35M of net debt. Fluence is also locked into buying $2.9B of battery cells from its suppliers, with up to $396M of penalties if it doesn't take the volumes—cells meant to feed the revenue plan it just cut.

CFO Ahmed Pasha said supporting FY2027 order growth may require another $300-500M of working capital over the coming year, against $863M of liquidity he expects back near the $900M level by fiscal year-end. Worth watching how that gets funded.

The new revenue guide is also not a low bar. Nine months of revenue totaled $1.59B, so the $3.0B midpoint requires ~$1.41B in the fourth quarter alone. That's 2.2x what Fluence just delivered, at the ~11% gross margin the guide implies, out of the same plants that caused the miss.

Management's case for hitting it, per the call, is that roughly half of what Q4 requires is already produced and fully integrated. The rest of the case:

At $13.20/share, the stock trades at 0.83x the revised guidance, against 0.78x when I wrote it up on the old guide. The bounce off the $10.55 low took back the whole discount, so the stock now costs more per dollar of guided sales than it did before the miss, on a guide that's $400M lower.

The open question is whether 5.9% was the cost of ramping new plants or the actual margin on the newer products, and the filing leans toward the products, tying the overruns to Gridstack Pro and Smartstack. Notably, this is the third straight quarter outside the target band (5.6% in Q1, 11.1% in Q2, 5.9% in Q3).

Management's answer is that the products aren't the problem, one-time costs are. The quarter carried ~$15M of new-product rollout and delay costs plus a $15M loss Fluence took on purpose on a battery supply agreement, most of it one project, to lock in long-term cell supply.

Nebreda's case is that FY2026 lands around the guided 12% excluding those one-time costs, and that backlog and new orders price in only the 10-15% range. That's a claim you should definitely remember when Q4 rolls around.

Energy Fuels (NYSE: UUUU), which I wrote up in June, reported Q2 revenue of $25.1M against $4.2M a year ago, and a net loss that widened to $33.4M. It's the largest US uranium producer, and it retrofitted its White Mesa Mill in Utah, the only operating conventional uranium mill in the country, to separate rare earths.

That widening loss isn't from the core business. Two charges explain it: $10.7M of transaction and integration costs for the pending ASM and VAC acquisitions, and $3.1M of accelerated reclamation accretion at the Kwale project in Kenya.

Underneath that, the uranium segment turned an operating profit for the first time, $271K on $25.1M of revenue against $24.8M of costs. The $271K itself is negligible, but the milestone isn't. Uranium is the segment meant to fund the rare earth buildout, and this is the first quarter it covered its own costs.

Production delivered. The mill made 865,000 finished pounds in Q2 and 1,655,000 in the first half, which already clears the low end of full-year guidance of 1.5-2.5M pounds at the halfway mark.

Two numbers in the mining detail are weaker. Pinyon Plain's grade fell to 0.71% in Q2 and 0.91% for the half, against ~1.62% in 2025. And mined pounds were 740,000 in the half against 2.0-2.5M guided for the year, so H2 needs a large step up.

The call filled in the cost side. Production cost came in at $23 per pound, which management calls an industry low, at the bottom of the $23-30 guided range, and finished-goods inventory cost fell to ~$33.92 per pound from ~$36 at the end of Q1. Management also expects grades to improve in H2 as mining reaches higher-grade zones at Pinyon Plain.

One thing the release doesn't make obvious is that the mill is now down for planned maintenance, with uranium processing set to resume in Q4 2026 or early 2027. Finished production pauses while the mining ramp continues.

In June I flagged five items to watch. This 10-Q updated four of them, and none of the four got better:

The fifth item is the only one that advanced. That's the heavy rare earth expansion, where construction began July 29 and the plan got bigger, up to 120 tonnes a year of dysprosium and 20 of terbium against the 48 and 14 in the original design, on ~$104M of capex.

The dilution is what you have to get comfortable with. Shares went from 240.4M at year-end to 249.9M at June 30, with $153.1M raised through the ATM.

From here, ASM adds ~15M shares at the end of August, VAC adds 65.9M shares plus preferred consideration and $718M of cash, the convertible adds another 34.4M above $20.34/share, and the OSC loan may add undetermined warrants.

The counterweights are $996M of working capital, a $250M standby loan from Goldman Sachs it hasn't touched, and a stated goal to “minimize dilution,” though the VAC close will consume a good chunk of the cash. CEO Ross Bhappu's answer was that "if it's accretive, I hate to think of it as dilution and think of it more as accretion."

At $13/share Energy Fuels is a ~$3.25B company, down 21.5% from the $16.56/share where I wrote it up in June. Net of ~$937M of cash and current marketable securities against the $700M convertible, enterprise value is ~$3.01B.

Framing the uranium business plus net cash at $2.0-2.5B leaves $0.75-1.25B for everything else. That's Phase 2, Vara Mada, the heavy circuits, ASM's metals platform, VAC's magnet business, and the medical isotope program.

Nothing this quarter said the assets are worth less. What changed is the price of owning them and the time until they pay off.

Celsius (Nasdaq: CELH), from my June write-up, posted record second quarter revenue of $817.9M, up 10.6% y/y, but net income fell 45% to $55.3M. The stock dropped 18% to about $24/share on the report, then rebounded to ~$27.80 the next day.

The rebound is its own ongoing story. Russ Weiner, who founded Rockstar and sold it to Pepsi for $3.85B in 2020, disclosed a 12M-share stake, ~4.7% of the company. He argued that management should be replaced, and volunteered himself for the CEO job. I’m not convinced he’d run the business any better than CEO Fieldly.

Celsius sells roughly one in five energy drinks bought in the US across three brands, CELSIUS, Alani Nu, and Rockstar, nearly all of it moving through PepsiCo's (PEP) trucks.

Most of that 45% drop is due to Celsius booking $80.9M of distributor termination fees to move more Alani Nu territory into Pepsi's system. Operating income fell to $75.3M from $143.0M, and the fee more than accounts for the decline. Back it out and operating income was $156.1M, up 9.2%.

Under the Channel Transition Amendment signed May 21, Pepsi reimbursed Celsius $81.1M for those exact fees, and Celsius had spent substantially all of it by June 30.

The catch is the accounting. Celsius expenses the fee now and books the reimbursement as deferred revenue amortized over the remaining ~16 years of the distribution agreement.

So the headline earnings decline is a timing artifact, and it's the second year running.

The bigger issue is margin. Q2 gross margin was 48.1%, down 340bps y/y, but last year's quarter carried a $21.7M inventory step-up charge that held it down to 51.5%. Strip that out and gross margin was 54.4% a year ago, so the underlying fall is closer to 630bps.

Adjusted EBITDA fell 12.4% to $184.2M even excluding the termination charge, and adjusted EBITDA margin went to 22.5% from 28.4%. Adjusted SG&A rose to 28.6% of revenue from 28.1%.

Promotional spending is what took those 630 basis points, and Celsius discloses it in a footnote rather than in the release. Promotional allowances deducted from revenue were $312.0M this quarter against $189.7M a year ago, up 64.5%.

Add those allowances back and gross billings grew ~21.6% while reported revenue grew 10.6%. Of the $201M of gross growth, $122M went straight back out in promotions.

Management split that spend in two on the call:

On the margin itself, CFO Jarrod Langhans called 48.1% in line with expectations, pointed at aluminum inflation, and guided Q3 gross margin to stay in the high-40s, adding that at current diesel and aluminum levels "margin expansion is largely offset."

The same promotional spending explains the brands. CELSIUS revenue fell 11.7% to $387.0M while retail scanner sales fell only 2%, so the gap is promotion, club-channel softness, and shipment timing, not consumers leaving. Management put the depletions-versus-orders timing at roughly half of it.

CEO John Fieldly took part of the blame himself, saying "we went too deep on the Celsius rationalization" and that he "would've not cut as many SKUs," with the 16-ounce line the weak spot and a replacement planned for 2027. He expects brand CELSIUS in Q3 to look a lot like Q2 in dollars, then exit the year back in growth.

Alani Nu revenue grew 21.0% to $364.4M while its scanner sales grew 55.7%, the 35-point difference being promotional allowances plus the shift into direct-store delivery. Rockstar added $66.5M with retail sales down 13%.

So Alani Nu is still growing faster than anything else Celsius owns, it's just converting far less of that growth into reported revenue than its scanner sales suggest.

Two call details sharpen that:

Outside of core operations, two legal items saw progress—both of which I highlighted in my June write-up:

The Flo Rida judgment is a check Celsius can write against $631.2M of cash. Beyond the direct financial hit, the real risk is how these investigations might impact Alani Nu's growth, a brand built on influencer marketing to a young, mostly female demographic.

On valuation, the balance sheet is the strong part of the quarter. Cash rose to $631.2M, net debt is only ~$37M, and Celsius repurchased ~$100M of stock in the quarter, ~$124M in the first half, against the $300M authorization it says it will keep using.

But the market cap understates what you're paying.

PepsiCo holds $1.76B of convertible preferred that ranks ahead of common, takes a 5% dividend first, and is why $55.3M of net income became $36.4M attributable to common.

Add the preferred and net debt to the ~$7.0B market cap and enterprise value is ~$8.8B, or ~12x annualized adjusted EBITDA, against Monster (MNST) at 30.3x trailing. Celsius trades at ~40% of Monster's EBITDA multiple.

The 10-point margin gap justifies part of that discount. The rest is the market betting Celsius never closes it.

Doximity (NYSE: DOCS) closed Friday at $27.40/share, up 32.6% on the day. Volume was 63.6M shares against a normal day of 2-3M.

When I wrote Doximity up in May at $19.20/share, I flagged what the Q1 FY27 report needed to show: (1) revenue at or above $152M, and (2) commentary on the AI Search pipeline.

Both landed. Revenue came in at $156.6M, up 7.0% y/y and above the $151-152M guide, and more than two dozen pharma programs are now contracted on AI Search.

Here's what the market is currently paying for:

Management also claims AI Search already earns more than 10x per search what it costs to run. And it repurchased $92M of stock in the quarter, with ~$400M left on the authorization against $688M of cash and no debt.

The AI headline was Doximity Ask ranking first among US-based models on the real-world portion of the NOHARM benchmark, which ran 1,100 clinical scenarios across 10 specialties, ahead of GPT-5.6 Sol, Claude Fable 5, and OpenEvidence. It's still a preprint, and OpenEvidence's CEO is disputing the methodology.

I don't expect Doximity to sustain its 32.6% jump for much longer, given the market likely overreacted to a quarter that cleared a low bar.

Disclaimer: This write-up is for informational and educational purposes only and is not investment advice. I may hold positions in the securities discussed. Do your own research before buying or selling any security.