Profitability Framework

Nail the objective and the baseline numbers before drawing anything

Two things need pinning down: what fell, and over what period. This is not a one-year shock with a single cause; it is a three-to-four-year structural slide, roughly $26 billion of revenue and about 20 percentage points of gross margin. That immediately tells you not to go hunting for a one-off event like a bad quarter or a customer loss. State the objective as the client would: restore double-digit operating margin, and decide whether the manufacturing business is worth keeping. Write down the two anchor numbers — revenue down about a third, gross margin down about 20 points — because every later calculation is checked against them.

Choose a revenue model that matches a company that owns its own factories

Price times Quantity is useless across thousands of SKUs from laptop chips to server CPUs. The right model has two layers. Revenue equals units shipped times average selling price, split by segment: client PC chips, data-centre chips, and external foundry work. But the cost model has to be wafer-driven: wafers started times die per wafer times yield, against a fab cost base that is overwhelmingly fixed. Saying this out loud sets up the whole case, because in semiconductors gross margin is mostly a story about how full the fabs are, not about what you paid for raw silicon.

Isolate the branch — and find that both branches are bleeding

Revenue fell about $26 billion, which at a 55 percent gross margin is roughly $14 billion of lost gross profit on its own. Separately, gross margin percentage fell about 20 points, which on the current $53 billion base costs another $10 billion or so. So this is not a revenue case or a cost case; it is a revenue case that mechanically caused a cost case. That link is the whole insight: fewer wafers through a fab whose depreciation, cleanroom, and staffing do not shrink means the fixed cost per unit rises. A fixed-cost-heavy business gets punished twice by a volume decline — once in lost contribution, once in unabsorbed overhead.

Build the cost side with real semiconductor line items

Do not leave two boxes labelled fixed and variable. The real lines are: wafer and materials cost (semi-variable), fab depreciation on roughly $25 billion a year of capex (fixed and rising), start-up costs for each new process node before it produces sellable volume (fixed and painful), R&D at roughly $16 billion a year, plus SG&A. Note the direction of travel: Intel was building five new nodes in four years and multiple new fabs, so depreciation and pre-production start-up costs were climbing at exactly the moment volume was falling. A candidate who names start-up costs and depreciation, rather than saying 'costs went up', is demonstrating industry knowledge.

Segment until the average breaks — Products versus Foundry

The company-level margin is a blended average that hides two completely different businesses. Once Intel began reporting Foundry separately, the picture was stark: in 2024, Intel Products did roughly $47-49 billion of revenue and still earned around $11-12 billion of operating income, a perfectly respectable design business. Intel Foundry did about $17.5 billion of revenue (mostly internal transfers) and lost roughly $13 billion. Cut Products further and the damage concentrates again: the PC segment was soft post-Covid, but the data-centre segment is where share was genuinely lost, as AMD moved from single-digit to roughly a quarter of server CPU units and, more importantly, data-centre budgets shifted from CPUs to Nvidia GPUs. So: one loss-making manufacturing arm, plus one structurally shrinking end-market inside an otherwise profitable products arm.

Check whether it is the firm or the industry

This is the step most candidates skip, and here it flips the whole recommendation. Over the same period TSMC's gross margin stayed in the 50s, Nvidia's ran around 75 percent, and AMD was gaining share — the industry was up, not down. Therefore the cause is internal, not an input-cost shock or a sector-wide price war. Trace it back and the root cause is a process-technology delay: missing the 10nm and 7nm timelines meant losing product competitiveness, which lost volume, which left expensive fabs underloaded, which crushed gross margin. One internal technical failure propagates through the entire profit tree — that is what a root cause looks like.

Quantify the leak, size the fix, and stress-test it

The two sized levers match the two segments. On Foundry, roughly $13 billion of annual operating loss can only be closed by filling the fabs with external customers or by shrinking the fab footprint — which is precisely why Intel separated Foundry into a subsidiary, courted external customers, cut capex, and slowed or shelved fab projects. On the fixed-cost base generally, the announced roughly 15 percent headcount reduction, a roughly $10 billion cost-reduction target, and suspending the dividend are all about buying time. The stress test is uncomfortable and worth saying aloud: the capex you cut to fix this year's margin is the same capex that funds the process node that restores competitiveness in three years. Cutting fixed cost in a technology race can be the move that guarantees you keep losing.

Nail the objective and the baseline before touching the tree

Confirm the shape of the decline: PAT down about Rs 120 crore, which at a 25 percent tax rate is a pre-tax gap of roughly Rs 160 crore, over two years, while AUM grew about 24 percent. Confirm there is no one-off — no fraud write-off, no merger, no accounting change. Confirm the metric the promoter actually cares about: he is talking in absolute profit, but the right lens for a lender is return on assets, and on these numbers RoA has fallen from roughly 4.0 percent to roughly 1.1 percent. Say that out loud early; it reframes the case from 'we lost Rs 120 crore' to 'every rupee of book is now earning a quarter of what it did', which is a very different problem.

Throw away Price times Quantity — build a lender's profit tree

A lending business has no units and no price in the manufacturing sense, and a candidate who draws P times Q here loses the room in minute two. The right tree is: Profit before tax = Net interest income + Fee income − Operating expenses − Credit cost. Net interest income itself equals average AUM times net interest margin, where NIM is yield on advances minus cost of funds. Flag the simplification honestly — cost of funds strictly applies to borrowings, not the whole book, so treating NIM as a spread on AUM is a working approximation. Now every line is something you can ask a number for, and each maps to a different real-world failure: pricing, funding, productivity, underwriting.

Bridge the Rs 160 crore pre-tax gap line by line

Ask for each driver in both years. Net interest income: 12.7 percent of Rs 4,500 crore is about Rs 572 crore in FY24, versus 9.8 percent of Rs 5,600 crore, about Rs 549 crore, so NII is down about Rs 23 crore despite a much bigger book. Fee and insurance-commission income fell from about 0.9 percent to about 0.4 percent of AUM (Rs 41 crore to Rs 22 crore), costing about Rs 18 crore after the client stopped aggressively bundling insurance. Operating expenses rose from 5.5 percent to 5.2 percent of AUM in ratio terms but from Rs 248 crore to Rs 291 crore in absolute terms, costing about Rs 44 crore. Credit cost went from 2.8 percent to 3.6 percent of AUM, Rs 126 crore to about Rs 202 crore, costing about Rs 76 crore. Total: about Rs 161 crore against a Rs 160 crore gap — the bridge closes, so you have found all the drivers and can now rank them.

Decompose the NIM line into a volume effect and a rate effect

NII barely moved in total, which naively looks harmless — but that is an average masking two large opposing forces, and this is where most candidates stop too early. Holding the old margin constant, growing the book by Rs 1,100 crore should have added about 12.7 percent of Rs 1,100 crore, roughly Rs 140 crore of extra NII. Instead, spread compression of 290 basis points on the full Rs 5,600 crore book destroyed about Rs 162 crore. So the growth the promoter is proud of did earn Rs 140 crore — it was simply handed straight back through pricing and funding. Splitting further: yield on advances fell from about 21.5 percent to 20.0 percent because of aggressive OEM-subvented schemes and competition from bank captives, while cost of funds rose from about 8.8 percent to 10.2 percent as the rate cycle turned and the client's credit rating stayed put.

Segment credit cost until the average breaks

Rs 202 crore of credit cost across a Rs 5,600 crore book averages 3.6 percent, which sounds survivable — but averages hide the problem. Ask for delinquency cut three ways: by sourcing channel (direct and repeat customers versus dealer-DSA sourced), by geography, and by disbursement vintage. The answer: loans written in the FY25 growth push through dealer-DSA channels in tier-3 UP and Bihar, at loan-to-value ratios above 90 percent and to first-time borrowers with no bureau history, are about 30 percent of AUM but generate about 70 percent of the losses — roughly 8 percent credit cost against about 1.7 percent for the rest of the book. Compute the contribution of that cohort properly: on Rs 1,700 crore it earns roughly 22 percent yield, but after about 10 percent cost of funds, about 8 percent credit cost and heavier collection opex of about 6 percent, it is earning close to nothing and consuming capital. The client grew its way into a loss.

Split the causes into external and internal — the recommendation depends on it

Test each driver against the industry. Cost of funds rising roughly 140 basis points is largely external — the rate cycle plus tighter bank lending to NBFCs hit every peer, so the recommendation there is about funding structure, not blame. Credit cost is the opposite: peer two-wheeler financiers were running roughly 2.5-3.0 percent, so the client's 3.6 percent blended and 8 percent on the problem cohort is firm-specific and therefore fixable. Yield compression sits in between: sector-wide competition is real, but the client chose to chase volume in the thinnest-margin schemes. That split matters because it tells you which levers are genuinely in management's control, and stops you recommending a cost-cutting exercise for a problem the whole market shares.

Recommend answer-first, with sized impact and named risks

Headline: the profit fall is not a growth problem, it is that roughly a third of the new book was underwritten at prices that never covered its risk — fix that cohort and about Rs 90-100 crore of pre-tax profit comes back. One, stop new dealer-DSA sourcing above 85 percent LTV for no-bureau customers and let that book run down by half; at roughly 8 percent credit cost on Rs 850 crore, that alone saves about Rs 65 crore a year and frees capital. Two, reprice what you retain in that segment by 150-200 basis points, worth roughly Rs 25-30 crore, since these borrowers are rate-insensitive and payment-EMI-sensitive. Three, attack cost of funds with co-lending and direct assignment with a bank partner on the prime slice of the book — moving 20 percent of AUM off balance sheet at a 150 basis point funding advantage is worth roughly Rs 17 crore, plus capital relief. Four, cut collection opex by pushing e-NACH and UPI autopay penetration, since field collection cost per case is what drove opex up in absolute terms. Risks to name: dealers may push volume to a competitor if you tighten, so protect relationships with faster sanction turnaround instead of loose credit; and shrinking the book raises the fixed-cost-per-rupee-of-AUM ratio, so the opex action has to run alongside the de-risking, not after it.