measured from the disclosures, not quoted from a brochure — each links to how it was done
P5Cost, measured
Regular plan costs 0.75% a year more than Direct
₹25,215 on ₹1 lakh over ten years. Same portfolio, two prices — the difference is what the Regular plan's distribution costs you, read from the two NAVs.
the factsheet archive starts Jun 2026, so this is a floor. A return earned before they arrived is the house's record, not theirs — the tenure table shows the fund against its category over exactly their months.
NAV sits 19 bp from its disclosed portfolio, but the replication is noisy
7 jumps since the last disclosure. We re-price the last disclosed portfolio every day and compare it with the NAV the fund published; a gap that opens and stays is a trade not yet disclosed.
Beat its category in 63% of three-year stretches, and averaged +2.4 points a year across all of them
109 rolling windows since 2014 · ahead by 5.0 points a year when it won, behind by 2.2 when it lost. A single five-year figure depends on the day you look; the share of stretches won does not — and the average margin says whether winning often was worth anything.
headline and record disagree One check found the headline number and the underlying record pointing different ways — marked on the cards below.
NAV and drawdown
Direct plan, growth class · as of 21 Sep 2026
Month-end NAV, indexed to 100 at Sep 2014
145 month-ends · ₹10.23 → ₹53.90, 5.3× since Sep 2014
Deepest fall (max drawdown)
−23.2%
23 Sep 2024 → 3 Mar 2025
Worst month
−11.4%
Mar 2026
Days to recover
not yet recovered
from the deepest trough back to the old high
Drawdown is measured on the last year of daily NAV; trailing returns are CAGR of the direct growth class. A single end date decides them — the consistency card (P4) below is the record that does not, and it reports both how often the fund beat its category and by how much, because the first without the second can point the wrong way.
Holdings
Aug 2026 disclosure · 46 lines · when each was first held and the change over the last three disclosures
Showing 21–30 of 46 · page 3 of 5rows per page102550all
Largest sectors, Aug 2026 · grey: a year ago
Petroleum Products11.6% · 10.3%
Electrical Equipment11.4% · 4.9%
Cement & Cement Products10.0% · 14.9%
Construction9.2% · 13.5%
Power8.8% · 5.6%
Telecom - Services8.1% · 8.1%
Industrial Products5.8% · 6.3%
Realty4.1% · 2.8%
share of the book05%10%15%
By market cap, Aug 2026
Large cap57.5%
Mid cap20.3%
Small / micro cap16.5%
Cash & equivalents2.1%
Not classified2.5%
Other1.1%
share of the book025%50%75%100%
"Not classified" is what no cap tier could be inferred for — newly listed names, or lines without an ISIN.
Composition by market cap, month by month — what the fund actually held against what its label says
Large cap 57%
Mid cap 20%
Small / micro 17%
Cash & other 2%
The seven checks, in full
each: the plain claim, where it sits in the category, and — folded — how it is measured and the case against. A flag fires only where headline and record disagree.
P4Consistency, not a single end dateheadline and record disagree
Beat its category in 63% of three-year stretches, and averaged +2.4 points a year across all of them
109 rolling windows since 2014 · ahead by 5.0 points a year when it won, behind by 2.2 when it lost
What this means. Take every three-year stretch since the fund had a NAV, one starting each month, and ask two things of each: did it beat the typical fund in its category, and by how much. Counting the wins tells you the record does not depend on when you happen to look; averaging the margin tells you whether winning often was worth anything, because four small wins do not pay for one large loss.
How often, and by how much. The count and the average agree. Ahead by 5.0 points a year in the 69 windows it won and behind by 2.2 in the 40 it lost, so the average across all 109 is +2.4 points. The worst window ended Dec 2020, 5.7 points behind.
63%
lowest in categoryhighest in category
62th percentile · above median of 138 funds · category median 56%
windows measured109windows won69average across every window+2.35 pts a year · median +2.52when ahead, by how much+5.02 pts a year over 69 windowswhen behind, by how much−2.25 pts a year over 40 windowsworst window−5.68 pts a year, ended Dec 2020best window+12.57 pts a year, ended Jan 2024non-overlapping windows in that span4 — the rolling count overlaps, this does not
Fund minus category, each rolling three-year window
ahead of categorybehind
Each bar is one window's margin in percentage points a year. The height of the bars, not the count of green ones, is what the average across every window measures.
A modest five-year number over a record that beat its category 63% of the time. Rolling every three-year window forward month by month gives hundreds of overlapping observations instead of one. This fund was ahead of its category median in 63% of them, which points the opposite way to the headline figure. A single end date decided that headline.
How it is measured, and the case against
A fund that beats its category in most rolling three-year windows has a record; one that beats it in a few has an end date. How often it won and by how much are different questions, and this card answers both, because a fund can win four windows in five and still be behind across all five.
How often has it beaten its category, across every window?
Why this might matterRolling every three-year window forward month by month gives hundreds of overlapping observations instead of one. A single end date decides a headline five-year figure; the share of windows won cannot be flattered by when you happen to look. The average margin across every window is that share and the payoff multiplied together — win share x average beat plus loss share x average shortfall is exactly the average — so a high share bought with a few ruinous windows cannot hide behind the count.
Why it might notRolling windows overlap heavily, so they are not independent observations and the true sample is far smaller than the count suggests.
P5Cost, measured
Regular plan costs 0.75% a year more than Direct
₹25,215 on ₹1 lakh over ten years
What this means. The Regular and Direct plans hold the same portfolio; the Regular one pays a distributor out of your money every day. The gap between the two NAVs is that cost, measured from the NAVs rather than quoted from a document.
0.75 pp a year
costliest in categorycheapest in category
95th percentile · top decile of 226 funds · category median 1.46 pp
on ₹1,00,000 over ten years₹25,215Direct vs Regular, annualised14.9% vs 14.1%measured over12 years
How it is measured, and the case against
The Direct plan is the same portfolio, priced differently; the divergence of the two NAVs is the cost, measured rather than quoted.
What does the Regular plan actually cost against its Direct twin?
Why this might matterSame manager, same holdings, same day. The gap between the two NAV series is the one number on this page that is knowable today rather than hoped for, and it compounds for as long as the units are held.
Why it might notA Regular plan buys advice. If that advice stops someone selling in a drawdown it can be worth more than the fee — the cost is visible and the behavioural saving is not, which makes this comparison unfair in exactly one direction.
P1Churn, and what it bought
Turns over 87% of the portfolio a year
−0.03 pts of excess return per unit of turnover
What this means. How much of the portfolio was bought and sold in a year, computed from what the fund disclosed each month. Trading costs something certain; the extra return it is meant to buy is not.
87% a year
costliest in categorycheapest in category
47th percentile · below median of 214 funds · category median 83%
excess return per unit of turnover−0.03 pts
How it is measured, and the case against
Turnover measured from consecutive disclosures — Σ|Δweight|/2 — costs something certain; the excess return it is meant to buy is not.
Is the trading paying for itself?
Why this might matterAMCs define their own turnover figure inconsistently. This one is computed from what the fund disclosed month to month, so two funds are comparable on it. Read it beside the excess return per unit of turnover.
Why it might notTurnover computed from month-end snapshots misses everything bought and sold inside a month, and a fund forced to sell by redemptions is charged for churn it did not choose.
P2Conviction or inertia
Typical top-10 holding kept 55 months
What this means. How long a typical top-10 position has been in the top ten. Long-held positions are a decision; a top ten that changes every quarter is a reaction.
55 months
lowest in categoryhighest in category
87th percentile · top quartile of 215 funds · category median 23 months
active sharenot yet computable — needs index constituent files
How it is measured, and the case against
The median months a top-10 position has been held separates a portfolio somebody chose from one that drifted.
How long does a top-10 position stay a top-10 position?
Why this might matterConcentration is how a manager expresses conviction, and persistence is how long they keep it. Long-held positions with a rising weight are a decision; a top ten that turns over every quarter is a reaction.
Why it might notPersistence without active share is inertia: a portfolio of 166 names that tracks its category closely has diversified away the reason to pay for it. Active share needs index constituent files, which are not loaded yet.
P3Hit rate of the top ten
5 of 10 top picks beat their peers over the next 6 months, and averaged 4.2 points ahead of them
628 positions judged, one disclosure at a time · ahead by 20.2 points when it won, behind by 14.3 when it lost
What this means. Take each month's ten largest holdings and check whether each one beat the typical stock its category holds over the next six months, and by how much. A blended return can hide two wins and eight losses; a count of wins can hide one position that lost half its value.
How often, and by how much. The count and the average agree. Ahead by 20.2 points in the 336 positions it won and behind by 14.3 in the 292 it lost, so the average across all 628 is +4.2 points. The worst position was INE397D01024 at the May 2020 disclosure, 55.9 points behind.
54%
lowest in categoryhighest in category
55th percentile · above median of 209 funds
positions judged628beat the median stock336average across every position+4.15 pts · median +1.70when ahead, by how much+20.23 pts over 336 positionswhen behind, by how much−14.35 pts over 292 positionsworst position−55.91 pts, INE397D01024 at the May 2020 disclosurebest position+179.08 pts, INE200A01026 at the Aug 2023 disclosure
How it is measured, and the case against
A manager's largest bets, judged one by one over the following six months, say more than the fund's blended return — and how far each one beat or missed the median stock says more than counting how many did.
Of its top-10 positions, how often did they beat the category-median stock?
Why this might matterA fund can beat its category on two positions and lose on eight; the hit rate shows the pattern the aggregate hides. The average margin then shows what the hit rate hides in turn: six winners worth a point each do not pay for four losers worth ten, and only the margin says which happened.
Why it might notSix months is short, sizing matters more than counting, and a manager who is right on the biggest position and wrong on nine small ones has done their job — the margins here are unweighted, so they answer the counting question better, not the sizing one. Overlapping disclosures judge the same position again each month it is held, so one long-held winner is counted many times. And this is a description of six months already past, not a forecast of the next six.
P6Size against edge
₹4,930 Cr, 87th percentile in category; 84% of growth came from inflows
smaller than 13% of the funds in its category (216 funds) · AUM Sep 2023 → Jul 2026 · Regular plan expense ratio 2.04%
What this means. Where the fund sits by size in its category, and how much of its growth was money arriving rather than the portfolio compounding. A very large fund cannot buy what a small one can; size should buy a lower fee in return.
87th percentile by AUM
largestsmallest
13th percentile · bottom quartile of 216 funds
median holding weight, trendfallingexpense ratio, Regular / Direct2.04% / 1.19% · category median 2.38%AUM, Sep 2023 → Jul 2026₹1,189 Cr → ₹4,930 Cr (+65% a year)of that change, from flows rather than returns84% net inflows · NAV +49% over the window
Assets under management, ₹ crore, Jun 2010 – Jul 2026
66 points · months without a factsheet figure use AMFI's quarterly average · Regular-plan expense ratio 2.69% in Jun 2018 → 2.04% in Sep 2026
How it is measured, and the case against
Past a certain size the smaller end of a mandate becomes unreachable, and the fund starts to look like its index; what size should buy in return is a lower expense ratio.
Has the fund outgrown the universe it invests in — and did size buy a lower fee?
Why this might matterSize narrows what the manager can meaningfully hold and should lower what the investor pays. Read the AUM history beside the expense ratio: a fund that has tripled through inflows while its Regular-plan TER sits above the category median has kept the benefit of scale for the house. The flow decomposition says how much of the growth was money arriving rather than the portfolio compounding.
Why it might notScale funds research teams, and plenty of large funds have gone on compounding for decades. Inflows follow performance, so a fund growing fast is usually one that did well; the constraint is real but not automatically binding, and a TER is a ceiling the AMC can cut at any time.
P7Whose record is this
Run by Bhavin Vithlani for at least 3 mo
the factsheet archive starts Jun 2026, so this is a floor
What this means. How long the people running it now have been running it, read from the monthly factsheets. A five-year record earned under someone else is not evidence about these managers.
No category distribution for this measure yet.
running it nowBhavin Vithlani (since at least Jun 2026)share of the fund's life under the current teamnot knowable — the archive starts Jun 2026, so the tenure is a floorchanges of hands in the archivenone
Who ran it, month by month · factsheets through Sep 2026
* dated by first appearance in the archive we hold, not by a date the factsheet printed. † already named on the first factsheet we hold: the stint is at least this long, and its start is not known. Co-managed months count for every manager named; a factsheet names the manager of record, not who made each call. All managers →
How it is measured, and the case against
A record earned under different people is not evidence about these ones.
Does the displayed record predate the people running it?
Why this might matterA five-year figure on a fund whose longest-serving manager joined two years ago is the house's record more than the person's. The tenure here comes from what each monthly factsheet named, so a change of hands is dated to the month it was disclosed, and the fund's return against its category can be read over exactly those months.
Why it might notFund houses have processes, research desks and mandates that outlast individuals, so a manager change is not a reset. Co-managed funds attribute the same months to every named manager, and a factsheet names the manager of record, not necessarily the person making the calls. And the archive we hold starts in Jun 2026: Bhavin Vithlani was already named on its first sheet, so the tenure here is the floor of what is known, not the month anyone arrived — the record may well belong to the people running it now.
Tracking gap
the NAV the fund published, against the NAV its last disclosed portfolio would have produced at each day's prices
NAV sits 19 bp from its disclosed portfolio, but the replication is noisy
What this means. Each day, the NAV the fund published against the NAV its last disclosed portfolio would have produced at that day's prices. A gap that opens and stays is a change the fund has made and not disclosed yet.
Latest gap beyond expense accrual
−19 bp
expense accrual ≈ 0.238 bp a day
Jumps since the disclosure
7noisy
30 Jul 2026 +32 bp · 17 Jul 2026 +29 bp · 8 Jul 2026 +26 bp
Window
60 days
25 Jun 2026 – 18 Sep 2026
Read it this way: daily residuals within ±2 bp are trading noise and the expense ratio accruing. A jump that stays is a portfolio change the fund has not disclosed yet — the size bounds how material. A gap that opens then closes is usually a dividend going ex-date.
Marked noisy: the replication jumped on 39 days in the window. Read the gap as coverage, not as a signal.
Share classes
latest NAV per AMFI code
Plan
Option
NAV
Date
direct SBI INFRASTRUCTURE FUND - DIRECT PLAN - GROWTH
growth
₹53.90
21 Sep 2026
direct SBI Infrastructure Fund - Direct Plan - Income Distribution cum Capital Withdrawal Option (IDCW)
idcw
₹47.71
21 Sep 2026
regular SBI INFRASTRUCTURE FUND - REGULAR PLAN - GROWTH
growth
₹49.41
21 Sep 2026
regular SBI Infrastructure Fund - Regular Plan - Income Distribution cum Capital Withdrawal Option (IDCW)
idcw
₹43.96
21 Sep 2026
The Direct / Regular gap, in rupees
Direct growth NAV
₹53.90
Regular growth NAV
₹49.41
NAV divergence to date
9.1% — same portfolio, priced differently
Regular costs more by
0.75% a year
On ₹1,00,000 over ten years
₹25,215
A Regular plan buys advice. The fee is visible and the behavioural saving is not; both are shown, neither is netted off.