Teardown

Digest · 2026-09-17

Scan #054: four US incumbents whose 2026 numbers admit the compounding has stopped — Mohawk peaked 5 years ago and paid $5B to enter the category (LVT) now cannibalising it, CenterPoint lost its Houston political franchise in the Beryl generator scandal, Sally Beauty is stabilizing on cost not growth as Ulta/Amazon/TikTok gut both segments, and Groupon has shed >80% of revenue since 2014 — against four emerging companies each betting a specific mechanism widens: Base Power's ERCOT gentailer + VPP stack now marked at $13B, Convex Insurance's 2019-vintage specialty book recapitalised at ~$7B by Onex + AIG in lieu of an IPO, Pactum's Walmart/Maersk/Vodafone reference deployments against SAP Ariba Joule and Coupa Compose shipping native negotiation agents, and Solvento's Mexico-side USMCA payables rails against TriumphPay's April 2024 peso switch-on

Eight companies. Four US incumbents where the compounding has visibly stopped — Mohawk Industries' revenue plateau + LVT self-cannibalisation, CenterPoint Energy's post-Beryl political-franchise reset, Sally Beauty's flat-to-+1% comps against Ulta/Amazon/TikTok on both segments, and Groupon's 80%+ revenue and headcount decline since 2014. Four emerging companies each betting a specific unbundling actually compounds: Base Power's Texas gentailer-plus-VPP now marked at $13B by Ribbit/Addition/Valor/JPMorganChase in August 2026, Convex Insurance's 2019-vintage specialty book recapitalised at ~$7B by Onex + AIG in February 2026 in lieu of the anticipated IPO, Pactum AI's Fortune 100 reference base against SAP Ariba Joule/Coupa Compose/Ivalua IVA Studio shipping native agent negotiation in 2026, and Solvento's Mexico-side USMCA freight-payables rails against TriumphPay's April 2024 peso switch-on.

The thread through today’s eight is that four US incumbents each ran the same playbook — buy your way into an adjacency, use the resulting scale to defend a legacy margin — and the playbook has now stopped working in a way the market has noticed. Mohawk paid $5B of M&A across Dal-Tile, Marazzi, IVC and Godfrey Hirst to enter tile and LVT, and LVT is now cannibalising the ceramic and hardwood books it bought. CenterPoint bought Vectren for $6B in 2019 for gas-utility scale and then paid the political price in Houston for the July 2024 Beryl blackout and the $800M mobile-generator scandal, ending up with a PUCT that trimmed its System Resiliency Plan from $5.75B to $2.7B. Sally Beauty is stabilizing FY2026 comps at 0-1% behind Fuel for Growth cost cuts while Ulta, Amazon, TikTok Shop and TikTok-native DTC brands attack Sally Beauty Supply’s DIY franchise and L’Oréal-owned SalonCentric outmuscles BSG on brand exclusivity. Groupon has shed >80% of revenue ($3.0B FY14 → $492.6M FY24), >70% of customers (55M+ 2015 → 15.4M FY24), and >80% of headcount, and Dusan Senkypl’s cost cuts have stabilised free cash flow without fixing the merchant value proposition. Against them: four emerging companies each betting a specific mechanism widens the wedge before the incumbent can bundle it or the market re-prices it — the questions are whether Base Power’s gentailer + VPP economics generalise outside ERCOT before the Series-D $13B mark demands proof, whether Convex’s 2019-vintage specialty book holds sub-95% combined ratios through softening now that the IPO route has been replaced by the Onex + AIG recap, whether Pactum’s Fortune 100 reference base compounds before SAP Ariba Joule and Coupa Compose ship native negotiation, and whether Solvento’s Mexico-side USMCA payables compounds before TriumphPay’s peso rails or a Nuvocargo/Cargado bundle takes the shipper AP integration back.

Mohawk Industries — Construction / Flooring · Incumbent, at risk. NYSE: MHK. The world’s largest flooring manufacturer, ~$11B FY2025 revenue across carpet, ceramic tile, laminate, hardwood and LVT, built through ~$5B of M&A (Dal-Tile 2002 ~$1.7B, Unilin 2005 ~$2.6B, Marazzi 2013 ~$1.5B, IVC 2015 ~$1.19B, Godfrey Hirst 2018 ~$400M). Revenue peaked in FY2021 at $11.74B and sits ~6% lower at $10.99B FY2025 despite a decade of bolt-ons, the market values the entire manufacturing base at under 0.75x sales, and Mohawk recorded an $876M goodwill impairment in Q3 2023 followed by ~$270M of restructuring charges. The uncomfortable finding: LVT — the category Mohawk paid IVC ~$1.19B in 2015 to enter — is now cannibalising the same ceramic and hardwood books Mohawk paid Dal-Tile and Columbia Forest Products billions to own. The attack surface is not another integrated flooring conglomerate; it is a China + Vietnam LVT direct-import brand plus a Shopify-native installer network that skips both the distributor and the big-box shelf.

CenterPoint Energy — Energy · Incumbent, at risk. NYSE: CNP. Houston-headquartered regulated electric and natural gas utility, ~$8.6B revenue FY2024, ~2.8M electric customers around Houston plus gas distribution in six other states. The at-risk verdict is not about long-run demand — it is about the July 2024 Beryl blackout (2M+ customers out up to 12 days), the $800M mobile-generator scandal (Life Cycle Power contract for gens that never deployed for Beryl), the resulting Texas PUCT that trimmed the System Resiliency Plan from $5.75B to $2.7B, and a Jason Wells CEO tenure that started January 2024 and inherited the political fallout. The Beryl aftermath handed the two most valuable customer classes — direct-connect AI data-center load and single-family residential — to ERCOT-native attackers (Base Power, on this same digest, is a direct beneficiary), and no rate-case win reverses the political reset in Houston.

Sally Beauty Holdings — Retail · Incumbent, at risk. NYSE: SBH. Denton, TX. Two segments — Sally Beauty Supply (~2,500 DIY retail stores) and BSG/CosmoProf (~1,300 professional-only stores, exclusive licensed-cosmetologist channel). ~$3.7B FY2024 revenue, ~22,000 employees, market cap oscillating between $1.0-1.3B. The recent stock rally reflects the July 2025 S&P ‘BB’ rating upgrade and Fuel for Growth cost cuts — not organic growth. FY2025 comps landed at +0.4%, FY2026 guidance is flat-to-+1%, and both segments face structural attackers: Ulta, Amazon and TikTok-native DTC brands (K18, dae, Amika, Curlsmith) gut Sally Beauty Supply’s DIY franchise from the top, while L’Oréal-owned SalonCentric — with exclusive access to Redken, Matrix and Kérastase — outmuscles BSG/CosmoProf on the brand exclusivity that was the professional channel’s whole moat. Denise Paulonis has 350+ store closures behind her since 2022 and is out of easy fat to trim.

Groupon — Ecommerce · Incumbent, at risk. NASDAQ: GRPN. Chicago. Local-commerce marketplace, once a $16B IPO in November 2011, now a ~$767M market cap (September 2026) with Dusan Senkypl (Pale Fire Capital, ~22% stake) as permanent CEO since May 2024. Revenue has fallen from ~$3.0B in 2014 to $492.6M FY2024, active customers from 55M+ in 2015 to 15.4M FY24, headcount from 11,000+ to under 2,000 (with another 400 — ~24% — cut in August 2026 in an AI-restructuring). Free cash flow has stabilised and short interest sits at ~57% (early 2025). The uncomfortable finding is that Senkypl’s turnaround is a cost story, not a demand one: the 2011 Posies Cafe TechCrunch post-mortem calling Groupon “the single worst decision I have ever made” describes a merchant value proposition — adverse selection, brand devaluation, no repeat-customer economics — that has not been fixed in fifteen years, and TikTok Shop, Booking Experiences, Toast Local and Yelp are all better-funded routes to the same local-demand pool.

Base Power Company — Energy · Emerging. Austin, TX. Residential battery-as-utility service on the ERCOT deregulated grid. Founded 2023 by Zach Dell (ex-SpaceX, Michael Dell’s son) and Justin Lopas (ex-SpaceX Starship manufacturing, ex-Anduril operations). Customers get an LFP battery installed for $0-595 upfront, buy retail electricity from Base at ~8.5¢/kWh + $19-29/month, and Base earns from the retail markup (wholesale at 3-5¢/kWh) plus ERCOT ancillary-service VPP dispatch. Recent cap-table: $68M Series A (2024, a16z + Valor + Trust Ventures), $200M Series B (April 2025, Addition/a16z/Lightspeed/Valor at ~$841M post per Sacra), $1B Series C (October 2025, Addition-led, $4B post), $1B Series D (August 2026, Ribbit/Addition/Valor/JPMorganChase co-lead at $13B). Employees ~368 (Revelio, March 2026). The falsifiable open question is whether three coupled mechanisms hold at once — gentailer margin covering ~$3,000-5,000 pack costs, VPP revenue funding pack payback under 7-8 years despite ERCOT ancillary revenues collapsing ~85% between 2023-2025 per Modo Energy, and in-house Austin manufacturing landing installed costs below Tesla Powerwall — before Sunrun Storm Ready, Tesla Powerwall Direct and Renew Home’s 16 GW aggregation force commodity retail-supply margins. BBB refund and install-delay complaints are the softer part of the record.

Convex Insurance — Insurance · Emerging. London + Hamilton, Bermuda specialty (re)insurer. Founded April 2019 by Stephen Catlin (ex-Catlin Group founder, sold to XL for £2.79B) and Paul Brand (ex-XL Catlin CEO of Insurance) with a ~$1.7B founding capital raise from Onex Corporation + PSP Investments, topped up ~$1B in 2020. GWP scaled from ~$1.1B (2020) to ~$5.9B (2025); combined ratio moved 173% → ~87%. In April 2025 Convex launched Lloyd’s Syndicate 1984. In February 2026, Onex + AIG jointly injected ~$7B in a recapitalisation that replaced the anticipated IPO, took PSP out, and installed Onex CEO Bobby Le Blanc as chair after Catlin’s March 2026 step-down. The mechanism question: does the 2019-vintage book hold sub-95% combined ratios through the 2024-2026 softening cycle, or does the greenfield actuarial cold-start show up as reserve strengthening on the 2019-2021 vintages once pricing gives up its 2023 peak? The recap-in-lieu-of-IPO is the tell that the public market was unwilling to pay the multiple Onex wanted on that book.

Pactum AI — Supply chain · Emerging. Mountain View, CA + Tallinn, Estonia. AI-native autonomous negotiation software for procurement teams — LLM agents negotiate commercial terms with tail-spend suppliers using category-specific playbooks. Founded 2019 by Martin Rand (CEO), Kristjan Korjus (CPO) and Kaspar Korjus (former Estonia e-Residency MD), seeded by Jaan Tallinn. Series A led by Atomico + General Catalyst; additional rounds through 2024-2025 with cumulative funding ~$50M+. Named Fortune 100 references include Walmart, Maersk, Vodafone, Wesco and Henkel. The mechanism question: does the reference base + Contract Space / Value Function scaffolding compound into a defensible data + workflow moat before (a) SAP Ariba Joule (GA Q1 2026), Coupa Compose with 20+ agents, and Ivalua IVA Studio ship native negotiation inside the source-to-pay suite the customer already pays for; (b) Zip HQ ($6B in reported customer savings by Dec 2025 with a native Price Negotiation Agent) owns the intake and buys negotiation as commodity; or (c) Ironclad/Sirion/Icertis ship negotiation as a free feature of the AI-CLM the legal team already owns.

Solvento — Logistics · Emerging. Mexico City. Cross-border USMCA freight-payables and factoring platform. Founded 2021 by Jaime Tabachnik (CEO), Pedro Bosch (CPO), and Guillermo Bosch (COO) — the founders in the earlier brief were incorrect and have been corrected. Funding: $5M seed October 2022 (Ironspring + Quona + Proeza + Dynamo + Zenda + Susa + 9Yards + Supply Chain Collective), $50M debt facility + $3.5M equity December 2023 (Lendable for debt; Quona for equity), $12.5M Series A November 2024 (Cometa-led). ~$71M cumulative. Employees ~50-100. The mechanism question: does Solvento Audita — the AI accounts-payable layer already live at Nestlé and 99minutos — embed deeply enough in shipper AP workflows that switching to TriumphPay (which switched on peso rails in April 2024) or a Nuvocargo/Cargado bundle costs the shipper more than the interchange it saves? And can the ~$180M cumulative invoice book compound into a proprietary carrier-credit dataset that Banxico’s falling policy rate (11.00% May-24 → 6.50% May-26) does not erase by making Konfio-style generalist SME credit cheaper than freight-specific factoring?

Full deep dives