This case raises questions about supplemental jurisdiction after removal, and about choice of law in a challenge to a limitation of liability in a waybill.

Siemens, a German manufacturer, made computer chips in Singapore, and sent them to San Jose, California, for testing. When the cartons containing the boxes of chips came back to Siemens' factory in Singapore, "one of the three inner boxes in each carton contained a brick instead of circuits."

Siemens had purchased insurance from Albingia Versicherungs, which paid Siemens about $235,000 for the stolen chips. Albingia brought this subrogation claim against all the firms in the shipping chain. Albingia sued in the California Superior Court for San Francisco County. The complaint stated five causes of action, the first arising under the Warsaw Convention, the other four arising under California state law for negligence, breach of contract, breach of the duty of care by bailee, and conversion. One of the defendants, Eva Air, an international air carrier, removed the case to federal court based on the Warsaw Convention claim.

After discovery, the case was settled as to all defendants but one, Schenker International, a freight forwarder. Schenker operated a warehouse in South San Francisco where the chips had been stored prior to transport on Eva Air back to Singapore. Schenker and Albingia stipulated to facts establishing, basically, that the chips had probably been stolen by Schenker employees while in Schenker's warehouse.

Siemens's waybills contained a limitation on liability supplementing the liability limitation in the Warsaw Convention. The contract said that, with exceptions immaterial to this case, liability where the Warsaw Convention did not apply was limited to $20 per kilogram. The waybill says that this amount is designed to be the approximate value of the "250 French gold francs" specified as the Warsaw Convention limit. Thus Schenker's idea was plainly, "our liability is limited to $20 per kilo if the Warsaw Convention does apply, and also if it doesn't."

Schenker's branch manager and warehouse manager watched a videotape from their closed-circuit security camera to figure out whether the shipment had been weighed when it came in. But somehow all the videotapes made from the security cameras during the whole four days the chips were in Schenker's care disappeared. The district court presumed that the chips were stolen by employees, for their own benefit and not for Schenker's benefit. This presumption has not been challenged in this appeal.

The parties filed cross motions for summary judgment on the applicability of the Warsaw Convention and the monetary limitations in the waybill. In its summary judgment motion, Albingia did not raise the issue of the jurisdiction of the federal court over the state law supplemental claims if the court found that the Warsaw Convention did not apply, nor the issue of the exercise of discretion by the court over retention of the supplemental claims. The district court determined that the Warsaw Convention did not apply, that the limitation on the waybill was valid under federal common law, and that Albingia was entitled to summary judgment in the amount of $5394, computed under the waybill limitations. Albingia appeals, to get the rest of the $235,000 plus expenses that it is out of pocket to its insured on the stolen shipment.

We review de novo the district court's grant of summary judgment.1 Albingia argues on appeal that once the district court held that the Warsaw Convention did not apply, the basis of removal — federal question jurisdiction — was gone, as was jurisdiction to consider the remaining state law claims. Albingia therefore argues that the district court should have remanded the state law claims back to the state Superior Court, because it lacked supplemental jurisdiction to rule on them. And Albingia further argues that even if the district court had supplemental jurisdiction, the district court should have applied California law rather than federal common law, and that under California law Albingia would have avoided the $20 per kilogram liability limitation.

Eva Air properly removed the case from state to federal court, because "[a]ny civil action of which the district courts have original jurisdiction founded on a claim or right arising under ... treaties or laws of the United States shall be removable...."2 No party filed a motion to remand upon removal. The complaint was drawn claiming a right to recover under a treaty and federal statute,3 the federal aspect was central to the claim and was not insubstantial or frivolous-for all Albingia knew, the chips were stolen while aboard Eva Air. Therefore, jurisdiction at the time of removal existed regardless of whether the Warsaw Convention ultimately turned out to supply a basis for recovery.4

There is also no serious question that in its summary judgment order, the district court first determined that the Warsaw Convention did not apply and then determined the state law limitations issue in favor of Schenker. We held in Read-Rite Corporation v. Burlington Air Express, Ltd.5 that where the loss to the goods occurs outside the airport, the Warsaw Convention does not apply.6 The Convention applies to transportation by air but not everything leading up to transportation by air.

Thus by the time the district court entered judgment in the case at bar, the federal question that justified removal had disappeared. The claim under the Warsaw Convention disappeared once the district court presumed and the parties stipulated that the bricks had been substituted for the chips in Schenker's South San Francisco warehouse.7

Albingia argues that once the district court correctly so decided, that should have been the end of the case in federal court. But instead of dismissing the state law claims or remanding them to state court, the district court exercised supplemental jurisdiction over them. Albingia contends that there was no supplemental jurisdiction over the remaining state law claims. As stated, supra, Albingia did not raise this issue until after the district court had ruled against it on its state law claims.

The supplemental jurisdiction statute provides that "in any civil action of which the district courts have original jurisdiction, the district courts shall have supplemental jurisdiction over all other claims that are so related to claims in the action within such original jurisdiction that they form part of the same case or controversy."8 In this case, the district court had a basis of original jurisdiction because of the federal question raised by Albingia's Warsaw Convention claim. Further, the state law claims plainly form part of the same case or controversy as the federal claim. Just like the Warsaw Convention claim, the state claims are four different theories for getting reimbursed for the $235,000 that Albingia paid Siemens for the stolen chips. The removal was without question proper at the time it was accomplished, because of the Warsaw Convention claim, so the district court plainly had supplemental jurisdiction at that time over the state law claims.9

The district court, in a single ruling on cross motions for summary judgment, held that because the chips were stolen in Schenker's off-airport warehouse, the Warsaw Convention did not apply, and the contractual limitation to $20 per kilogram was valid under federal common law. Albingia argues that upon deciding that the Warsaw Convention did not apply, the district court lost jurisdiction to decide that the contractual limitation applied. That contention is incorrect. Where removal was proper, but the federal question claim is defeated on the merits, the district court nevertheless may in its discretion retain supplemental jurisdiction over the state law claims.10 The Supreme Court held in Bell v. Hood that when a well-pleaded complaint states a federal claim, "[j]urisdiction, therefore, is not defeated... by the possibility that the averments might fail to state a cause of action on which the petitioners could actually recover," and "[i]f the court does later exercise its jurisdiction to determine that the allegations in the complaint do not state a ground for relief, then dismissal of the case would be on the merits, not for want of jurisdiction."11 In the case at bar, the federal claim was dismissed on the merits rather than for lack of jurisdiction, because it turned out that the chips were stolen at Schenker's off-airport warehouse. Jurisdiction existed because the pleading asserted, without frivolousness, that the Warsaw Convention gave Albingia a claim.

It is arguably an obstacle to this interpretation that the statute governing procedure after removal has subsequently been amended. In pertinent part, the removal statute, section 1447(c), provides that motions to remand for procedural defects must be made within 30 days of removal, but "[i]f at any time before final judgment it appears that the district court lacks subject matter jurisdiction, the case shall be remanded."12 Wright & Miller says that "[a]n interpretive debate has arisen whether the language `lacks subject matter jurisdiction' in the current text of Section 1447(c) refers only to defects existing at the time of removal or takes into account subsequent events."13 The words "any time" and "lacks," in the present tense, suggest mandatory remand, but the word "case" suggests that jurisdiction must be lacking over the entire case, not merely a claim within the case.

It was well-settled under the previous version of the removal statute14 that a federal district court retains "the power to hear claims that would not [have been] independently removable even after the basis for removal jurisdiction is dropped from the proceedings."15 Our sister circuits that have ruled on the question interpret the statute as amended not to deprive the district court of jurisdiction to decide the related state claims even though the federal claim upon which removal was based is dismissed on the merits.16 Thus, in our post-amendment en banc decision in Acri v. Varian Associates,17 we held that even if all federal claims are dismissed before trial, this "has never meant that they [state law claims] must be dismissed."18 This interpretation preserves coherence with the new supplemental jurisdiction statute, which provides that where the district court has original jurisdiction over a federal claim, it "shall have supplemental jurisdiction" over related state law claims.19

The supplemental jurisdiction statute, section 1367, says that district courts "shall have" jurisdiction over the non-federal claims forming part of the same case or controversy, which they "may decline to exercise" under denoted circumstances,20 while the procedure after removal statute, section 1447, says that "[i]f at any time before final judgment it appears that the district court lacks subject matter jurisdiction, the case shall be remanded."21 Here is how we reconcile the "may decline to exercise" language of the supplemental jurisdiction statute with the "shall be remanded" language of the procedure after removal statute: if state law claims are asserted as part of the same case or controversy with a federal claim, the district court has discretion to exercise supplemental jurisdiction over the remaining state law claims and the mandatory remand provision of the procedure after removal statute does not apply. Under the plain language of the statutes, logically it cannot "appear[] that the district court lacks jurisdiction" under 1447(c) if it "shall have" jurisdiction under 1367.

The Second Circuit held in Parker PPA v. Della Rocco, Jr.,22 that "§ 1447(c) merely addresses the consequences of a jurisdictional flaw, i.e. it mandates a remand rather than a dismissal."23 Parker relies on language in a Supreme Court decision, International Primate Protection League v. Administrators of Tulane Educational Fund,24 that, in somewhat different circumstances, reads the new language in section 1447(c) to mean that "a finding that removal was improper deprives that court of subject matter jurisdiction and obliges a remand."25 That is to say, section 1447(c) does not mean that if the federal claim drops out, the district court must remand; it means that if there is no jurisdiction — federal question, supplemental, diversity, or otherwise — the district court must remand the removed case rather than dismissing it. Our cases already bear this out; now we hold explicitly what, in Acri and other cases,26 we have held implicitly: section 1447(c) means that if it is discovered at any time in the litigation that there is no federal jurisdiction, a removed case must be remanded to the state court rather than dismissed. Section 1447(c) does not mean that if a facially valid claim giving rise to federal jurisdiction is dismissed, then supplemental jurisdiction is vitiated and the case must be remanded. Once supplemental jurisdiction exists, it remains, subject to the discretionary provision for remand in section 1441.

In the case at bar, because Albingia's state law complaint was properly removed to the district court due to the federal question asserted by Albingia's Warsaw Convention claim, the district court had supplemental jurisdiction over the state law claims. The district court's supplemental jurisdiction over the state law claims was not destroyed by dismissal of the Warsaw Convention claim. This is consistent with the long established approach to diversity cases, where even if the damages proved turn out to be less than the jurisdictional amount, the district court nevertheless retains jurisdiction to render a final judgment on the merits.27 Supplemental jurisdiction is not destroyed by elimination of the basis for original jurisdiction.

The only remaining issue relating to jurisdiction is whether, after the federal claim was dismissed, the district court abused its discretion by retaining the case rather than remanding it.28 Our en banc decision in Acri holds that although the district court "may," under section 1367(c), decline to exercise its supplemental jurisdiction in these circumstances, and ordinarily "should" dismiss the state law claims under United Mine Workers v. Gibbs, that "has never meant that they must be dismissed."29 Here, considering the Acri factors,30 there was plainly no abuse of discretion, because Albingia waited until the dealer turned up his hole card, in the court's summary judgment decision, before claiming that the adverse decision should be vacated because federal question jurisdiction had disappeared. We decline to let Albingia take its chips off the table because it didn't like the dealer's hand.

Albingia argues that, even if the district court did have jurisdiction and permissibly exercised it, the choice of law decision is erroneous. The parties disagree whether, under California law, the $20 per kilogram limit in the waybill is enforceable. Although the disagreement is arguable,31 we need not resolve the disagreement under California law. Albingia does not dispute that if federal common law applies, the limit is enforceable. We review de novo a district court's choice of law decision.32

Two prior Ninth Circuit decisions that came down three weeks apart control the determination of which law applies. Our decision in Read-Rite Corp. v. Burlington Air Express33 holds that where the Warsaw Convention is inapplicable because the damage occurred off-airport, federal common law applies to determine the validity of a waybill limitation on liability for damage to goods being shipped in interstate or international commerce by air.34 Under federal common law, the limit on liability is valid and enforceable if the shipper has reasonable notice of it and a fair opportunity to purchase the means to avoid it.35 Purchasing separate insurance on the cargo, as Siemens did from Albingia, satisfies both requirements.36 Read-Rite controls the case at bar and establishes, first, because Schenker's warehouse was outside the airport, the Warsaw Convention did not apply, and second, because Siemens bought insurance on the chips from Albingia, in obvious recognition of the waybill limitation, the $20 per kilogram limit is valid under controlling federal common law.

Our decision three weeks after Read-Rite in Insurance Company of North America v. Federal Express Corporation,37 concluded that state law applied. But it is distinguishable on its facts. In Federal Express, as in the case at bar, computer chips were stolen out of a warehouse by employees. The difference, though, is that in Federal Express the warehouse was within the airport. The Warsaw Convention applies once the goods are within the airport.38 In Federal Express we read Article 25 of the Warsaw Convention to refer us to state law to determine whether willful misconduct by a shipper's employee invalidates the limit on liability.39

Federal Express and Read-Rite come out differently because the Warsaw Convention applies in the former but not the latter. It may seem nonintuitive that state law controls once the shipment is at the airport because an international treaty governs the commercial relationships, while federal common law controls when the shipment is sitting within the state, is outside the airport at a South San Francisco warehouse and beyond the scope of the Warsaw Convention.40 But that's the way it is, under our controlling authority.

AFFIRMED.